{
 "about": "海外 VC / 投资人自己写的 AI agent 支付文章（投资备忘、赛道论点、市场地图、年度预测），外加投研机构与大公司联合发的行业报告；每篇打开原文核对过标题、日期、作者。stance = 对「稳定币 vs 卡」「现在有没有真实需求」等争议的立场。vs / why / rail / top / region 是我们的归类，不是文章原话：vs = 与图谱判断「agent 花钱、收钱都要补新的一层，现有支付默认有人在场」一致 / 部分 / 相反 / 无关，rail = 稳定币 vs 卡的立场，top = 最值得读。gist 是核心观点的中文转述，不是原文；seg 对应 layers.json 的 code，all = 赛道总论。",
 "articles": [
  {
   "t": "When AI Agents Start Spending Your Money | Alex Rampell & Terry Angelos（The a16z Show）",
   "org": "a16z",
   "au": "Alex Rampell（主持）、Terry Angelos（VGS）",
   "date": "2026-10-15",
   "url": "https://a16z.simplecast.com/episodes/when-ai-agents-start-spending-your-money-alex-rampell-terry-angelos-YowGFPLh",
   "kind": "赛道论点",
   "seg": [
    "a5",
    "g4",
    "k6"
   ],
   "co": [
    "VGS",
    "Stripe（Agentic Commerce / ACP）",
    "PayPal Agentic Commerce Services",
    "Shopify Agentic Storefronts"
   ],
   "gist": "agent 真买东西时，要给它的是保留拒付、身份核验、商户关系的「agentic token」，而不是裸卡号；PayPal、Shop Pay、Link 在抢成为 agent 默认钱包。节目说 VGS 每周已发放数百万个这类 token；AI 还可能让依赖用户不留意的订阅和费用「损耗」消失。",
   "cav": "播客节目，非书面文章；嘉宾 Terry Angelos 是 VGS 的执行董事会成员，a16z 是 VGS 投资人，利益相关。节目页日期 2026-10-15，搜索摘要有标 10-14。",
   "stance": "稳定币 vs 卡：主持人认为加密是「持有人工具」，没有拒付，商业上还是要 Visa/万事达在中间；需求：真实，已经在发 token。",
   "vs": "部分",
   "why": "要新的是 agent 令牌（保留拒付、身份），但跑在卡网络上，不需要新轨道",
   "rail": "卡",
   "top": 1,
   "region": "美国"
  },
  {
   "t": "What Is Agentic Identity? Meet Baselayer, the Identity Infrastructure for AI Agents",
   "org": "M13",
   "au": "Katie Boysen",
   "date": "2026-09-22",
   "url": "https://www.m13.co/article/what-is-agentic-identity-meet-baselayer-the-identity-infrastructure-for-ai-agents",
   "kind": "投资备忘",
   "seg": [
    "g4"
   ],
   "co": [
    "Baselayer"
   ],
   "gist": "AI 从给答案变成执行任务，需要新的安全、治理、身份和授权层；机构放行 agent 交易之前，必须知道是谁部署了它、这方可不可信、被授予了什么权限，M13 认为已有身份记录、风险洞察和银行关系的公司最适合做这层信任。",
   "cav": "Baselayer 3500 万美元 A 轮 M13 领投；偏身份/KYA，与支付相关但不是支付轨道。页面日期读到 2026-09-22（搜索摘要为 09-21）。",
   "stance": "需求：金融机构需要「知道你的 agent」才敢放行。",
   "vs": "一致",
   "why": "机构放行 agent 交易前先要知道是谁部署、授权了什么，这正是授权与风控层",
   "rail": "不涉及",
   "region": "美国"
  },
  {
   "t": "Agentic Commerce and Agentic Fintech: Where the Next Great Fintech Companies will be Built",
   "org": "645 Ventures",
   "au": "Dawit Wondimagegn、Nnamdi Okike",
   "date": "2026-07-21",
   "url": "https://www.645ventures.com/articles/agentic-commerce-and-agentic-fintech",
   "kind": "赛道论点",
   "seg": [
    "all",
    "m7",
    "g4"
   ],
   "co": [
    "Nekuda",
    "Skyfire",
    "Catena"
   ],
   "gist": "Agentic commerce 是六层栈，支付轨道、身份和协议被大公司占住，创业公司最有机会在中间件/编排和购后环节；近期价值在基础设施和 B2B 采购、后台自动化，「自动理财」是应用层最大机会，但依赖开放银行、即时结算等条件。",
   "cav": "645 Ventures 自己有 Nekuda、Skyfire、Sponge 等相关投资（文中标注为其组合公司）。",
   "stance": "稳定币 vs 卡：认为轨道/身份/协议被在位者占住；需求：近期在 B2B，消费端靠监管条件放开。",
   "vs": "部分",
   "why": "同意要新的一层，但认为轨道、身份、协议已被大公司占住，创业公司的机会在中间件和购后",
   "rail": "不涉及",
   "region": "美国"
  },
  {
   "t": "Investing in Natural: Money That Moves at the Speed of Software",
   "org": "Forerunner Ventures",
   "au": "Kirsten Green、Fawzi Itani",
   "date": "2026-07-20",
   "url": "https://www.forerunnerventures.com/perspectives/investing-in-natural-money-that-moves-at-the-speed-of-software",
   "kind": "投资备忘",
   "seg": [
    "w1",
    "r2",
    "g4"
   ],
   "co": [
    "Natural"
   ],
   "gist": "Agent 已经会计划和行动，但支付系统仍默认每笔交易由人发起；Forerunner 认为持久价值会集中在三处：记录钱在哪的账本、管 agent 权限与责任的信任与责任层、以及之后基于余额和策略数据做的信贷。",
   "cav": "领投 3000 万美元 A 轮；页面日期读到 2026-07-20（搜索摘要为 07-19）。",
   "stance": "需求：认为是基础设施缺口，不是已有大量交易。",
   "vs": "一致",
   "why": "支付默认有人发起，价值会落在账本、信任与责任层和之后的信贷",
   "rail": "不涉及",
   "top": 1,
   "region": "美国"
  },
  {
   "t": "Why We Invested in Natural: Payments for AI Agents",
   "org": "Torch Capital",
   "au": "Jake Moross、Chris Harper",
   "date": "2026-07-20",
   "url": "https://torchcapital.vc/blogs/why-we-invested-in-natural",
   "kind": "投资备忘",
   "seg": [
    "w1",
    "g4"
   ],
   "co": [
    "Natural"
   ],
   "gist": "现有支付默认有人在场，agent 需要新的身份核验、欺诈模型和可编程工具；Torch 看好 B2B 支付作为最先落地的市场，以及 Natural 把每笔付款关联到 agent 与指令的身份层、信贷优先的做法。",
   "cav": "Torch 是 Natural A 轮的跟投方之一。",
   "stance": "需求：B2B 先行。",
   "vs": "一致",
   "why": "agent 需要新的身份核验、欺诈模型和可编程工具，B2B 先落地",
   "rail": "不涉及",
   "region": "美国"
  },
  {
   "t": "Agentic Payments from the Ground Up: What the Onchain Data Tells Us",
   "org": "Visa × Artemis",
   "au": "Tim Conard（Visa）、Lucas Shin（Artemis）",
   "date": "2026-07",
   "url": "https://www.visa.com/en-us/thought-leadership/innovation/agentic-payments-from-the-ground-up",
   "kind": "行业报告",
   "seg": [
    "all",
    "p3",
    "s0"
   ],
   "co": [
    "Visa Intelligent Commerce",
    "Coinbase x402",
    "Stripe（Agentic Commerce / ACP）",
    "Tempo"
   ],
   "gist": "Agentic commerce 分两类：代表人类买机票、订阅的「宏观」，消费级金额，走现有卡轨道；软件对软件的亚美元高频「微观」，卡的固定费用算不过来，适合稳定币。预计两种轨道混合共存，Visa 要做两边都支持的基础设施。",
   "cav": "Visa 是卡组织，立场不中立；页面日期为「最后更新 2026-07-14」，数据截至 2026-04-21，PDF 报告里写 2026-07；Artemis 用自家方法剔除刷量。",
   "stance": "稳定币 vs 卡：并存（卡管宏观，稳定币管微观）；需求：x402 有持续的机器对机器流量，MPP 还很小。",
   "vs": "部分",
   "why": "替人买东西的大额交易照旧走卡；只有机器之间的亚美元小额要新轨道",
   "rail": "并存",
   "top": 1,
   "region": "美国"
  },
  {
   "t": "The Shift from Selling Software to Selling Work",
   "org": "Benhamou Global Ventures",
   "au": "Jessica（BGV，仅署名名字）",
   "date": "2026-06-24",
   "url": "https://www.benhamouglobalventures.com/news-and-insights/shift-from-selling-software-to-selling-work",
   "kind": "赛道论点",
   "seg": [
    "b9"
   ],
   "co": [],
   "gist": "AI 原生公司的定价从「卖软件使用权」转向「卖完成的工作」；文中引用 Kyle Poyar 对 230 多家公司的调查，混合定价（订阅加用量）从 25% 升到 37% 成为主流。按席位定价在 agent 干活时失效，用量/结果定价要求成效可归因、可衡量。",
   "cav": "作者页面只写名字 Jessica；数字转引自 Kyle Poyar（Growth Unhinged）的调查，非 BGV 自测。",
   "stance": "稳定币 vs 卡：不涉及；需求：定价模式迁移已经发生，结果定价依赖可衡量的成效。",
   "vs": "一致",
   "why": "按席位定价在 agent 干活时失效，计费要改成按用量、按结果",
   "rail": "不涉及",
   "region": "美国"
  },
  {
   "t": "Investing in ShopAgentic – Commerce, Powered by Agents",
   "org": "Greenfield Capital",
   "au": "Claude Donzé",
   "date": "2026-06-11",
   "url": "https://greenfieldcapital.com/2026/06/11/investing-in-shopagentic-commerce-powered-by-agents/",
   "kind": "投资备忘",
   "seg": [
    "m7"
   ],
   "co": [
    "ShopAgentic"
   ],
   "gist": "AI 助手正在接管选品和购买，但商户系统是为人类浏览器造的；ShopAgentic 在现有系统上加一层 agent 原生能力，让买方 agent 读得懂目录、价格、库存，协议无关的编排层有网络效应。Greenfield 还认为它以后可以从经过自家支付轨道的交易里分到价值。",
   "cav": "€190 万 pre-seed，与 May Ventures 共同领投；商户侧（非支付轨道）为主。",
   "stance": "需求：押注商户必须适配 agent 买家。",
   "vs": "一致",
   "why": "商户系统为人类浏览器造，要加一层让买方 agent 读得懂的结账能力",
   "rail": "不涉及",
   "region": "其他"
  },
  {
   "t": "Catena: The Financial Backbone for the New Economy",
   "org": "Acrew Capital",
   "au": "Lauren Kolodny、Maitree Mervana Parekh",
   "date": "2026-05-20",
   "url": "https://perspectives.acrewcapital.com/p/catena-the-financial-backbone-for",
   "kind": "投资备忘",
   "seg": [
    "w1",
    "g4",
    "s0"
   ],
   "co": [
    "Catena"
   ],
   "gist": "两个趋势在汇合：稳定币让美元成为互联网原生，AI 让互联网成为 agent 原生；现有金融系统撑不住机器速度的自主交易。Catena 做 AI 原生银行：给 agent 验证过的身份、账户和多轨支付，并给部署方策略、限额、监督的控制面；创始人是 Circle 里造 USDC 的人。",
   "cav": "Catena A 轮 3000 万美元，a16z crypto 与 Acrew 共同领投；Acrew 为该轮领投方。",
   "stance": "稳定币 vs 卡：站稳定币加受监管银行；需求：判断为趋势。",
   "vs": "一致",
   "why": "现有金融系统撑不住机器速度的交易，要给 agent 身份、账户和控制面",
   "rail": "稳定币",
   "region": "美国"
  },
  {
   "t": "Who Pays the Agent? The Race for Frictionless Machine Payments",
   "org": "Keyrock（与 Coinbase、Tempo、Virtuals 联合）",
   "au": "Ben Harvey",
   "date": "2026-05",
   "url": "https://keyrock.com/who-pays-the-agent/",
   "kind": "行业报告",
   "seg": [
    "p3",
    "s0",
    "n8"
   ],
   "co": [
    "Coinbase x402",
    "Stripe（Agentic Commerce / ACP）",
    "Tempo",
    "Virtuals Protocol"
   ],
   "gist": "四套机器支付架构（x402、MPP、AP2、Visa 卡轨道）其实在叠成一个栈，不是互相淘汰；亚美元小额付款卡的固定费用算不过来，稳定币占优。文中说 agent 已结算超过 7300 万美元、约 1.76 亿笔，98.6% 用 USDC，约 76% 的 x402 付款低于 0.30 美元。",
   "cav": "页面未写具体发布日（只能确认 2026-05）；Keyrock 是做市商，报告与 Coinbase、Tempo 合写，立场偏加密；文中同时提示对单一稳定币 USDC 依赖的集中风险。",
   "stance": "稳定币 vs 卡：小额支付稳定币赢，大额随信任基础设施成熟再上链；需求：已有但体量很小（文中 7300 万美元）。",
   "vs": "一致",
   "why": "本图谱的骨架：四套机器支付架构叠成一个新栈",
   "rail": "稳定币",
   "top": 1,
   "region": "欧洲"
  },
  {
   "t": "Agents are starting to operate real systems — who's actually in control?",
   "org": "a16z crypto",
   "au": "Christian Catalini、Andy Hall、Noah Levine、Christian Crowley",
   "date": "2026-04-18",
   "url": "https://a16zfrontier.substack.com/p/agents-are-starting-to-operate-real",
   "kind": "赛道论点",
   "seg": [
    "all",
    "g4",
    "p3",
    "s0"
   ],
   "co": [
    "Coinbase x402",
    "Stripe（Agentic Commerce / ACP）",
    "Tempo"
   ],
   "gist": "Agent 已成为经济参与者，但身份、可审计记录、结算这些基础设施没跟上，区块链能补；agent 需要「知道你的 agent」（KYA）凭证，相当于 agent 世界的 SSL，否则商户会继续把 agent 挡在门外。稳定币让任何开发者不用商户协议就能把接口变成可付款。",
   "cav": "文中说剔除刷量后 x402 每月约 160 万美元 agent 付款，远低于彭博报道的 2400 万美元。",
   "stance": "稳定币 vs 卡：稳定币适合无承保的「无头商户」；需求：文中自己承认真实量很小（约 160 万美元/月）。",
   "vs": "一致",
   "why": "身份、可审计记录、结算没跟上，要 KYA 凭证，否则商户继续挡 agent",
   "rail": "稳定币",
   "region": "美国"
  },
  {
   "t": "The Checkout is Dead pt2: Invisible Commerce",
   "org": "Fintech Brainfood",
   "au": "Simon Taylor",
   "date": "2026-04-05",
   "url": "https://www.fintechbrainfood.com/p/the-checkout-is-dead-pt2",
   "kind": "赛道论点",
   "seg": [
    "p3",
    "m7",
    "g4"
   ],
   "co": [
    "Tempo"
   ],
   "gist": "agent 把支付变成「不可见」：由现实事件触发（如按位置自动付停车费），不再有结账页；文中用沃尔玛在 ChatGPT 里结账转化率比自家网站低 66% 说明套用人类结账行不通。机器原生协议（MPP）更合适，但身份（KYA）、审计、责任要先解决。",
   "cav": "Simon Taylor 个人通讯，非基金官方；沃尔玛 66% 的数字是文中引用。",
   "stance": "稳定币 vs 卡：重点在协议和身份而非轨道；需求：聊天框内结账需求弱（沃尔玛数据），机器事件触发的需求强。",
   "vs": "一致",
   "why": "结账页消失，由事件触发付款；先要解决身份、审计和责任",
   "rail": "不涉及",
   "region": "欧洲"
  },
  {
   "t": "Machine Economy 2030",
   "org": "Artemis",
   "au": "Lucas Shin",
   "date": "2026-03-30",
   "url": "https://research.artemis.ai/p/machine-economy-2030",
   "kind": "行业报告",
   "seg": [
    "all",
    "n8",
    "s0"
   ],
   "co": [],
   "gist": "预测到 2030 年 agent 取代浏览器和店面，成为人用互联网的主要方式，产生持续不断的小额自动购买；现有支付和广告模式扛不住，需要微支付轨道和稳定币结算，按次付费替代广告。价值集中在入口界面、支付、算力三层，长尾 agent 商业走开放协议而不是平台目录。",
   "cav": "Artemis 是链上数据与研究公司（非传统 VC），收在行业报告；属预测，文中为粗略估算。",
   "stance": "稳定币 vs 卡：稳定币做微支付；需求：当前小，押注增长极快（用 ChatGPT 的普及速度类比）。",
   "vs": "一致",
   "why": "持续的小额自动购买需要微支付轨道，按次付费替代广告",
   "rail": "稳定币",
   "region": "美国"
  },
  {
   "t": "The Agentic Economy Will Be Massive. Agentic Commerce Won't",
   "org": "Dragonfly",
   "au": "Robbie Petersen",
   "date": "2026-03-23",
   "url": "https://writing.dragonfly.xyz/post/the-agentic-economy-will-be-massive-agentic-commerce-won-t",
   "kind": "赛道论点",
   "seg": [
    "all",
    "s0"
   ],
   "co": [],
   "gist": "大多数 agent 不会自主花钱：企业里的 agent 像 SaaS，成本被平台打包成一张账单（4 万次 API 调用只产生一张发票）；消费者 agent 只负责检索，最后由人授权付款。只有少数自下而上、真正自主的 agent 需要自己付款，加密轨道赢在无需许可，而不是更便宜。",
   "cav": "页面未写作者和日期；作者与日期（2026-03-23 前后）来自转载页（ChainCatcher/HTX）；作者其后（2026-06）加入 a16z crypto。",
   "stance": "稳定币 vs 卡：只在狭窄的自主 agent 场景加密赢，原因是无需许可；需求：多数 agent 不需要新支付轨道，「大部分按月结算」。",
   "vs": "相反",
   "why": "多数 agent 不自己花钱：企业 agent 的成本被平台打包成一张发票，消费者 agent 最后由人付款",
   "rail": "卡",
   "top": 1,
   "region": "其他"
  },
  {
   "t": "'Open Agentic Commerce' and the end of ads",
   "org": "a16z crypto",
   "au": "Samuel Ragsdale",
   "date": "2026-03-21",
   "url": "https://a16zfrontier.substack.com/p/open-agentic-commerce-and-the-end",
   "kind": "赛道论点",
   "seg": [
    "p3",
    "n8"
   ],
   "co": [
    "Merit Systems"
   ],
   "gist": "ChatGPT/Gemini 里的 ACP、UCP 结账只是渐进改良；真正的大变化是开放协议（x402、MPP）加稳定币，让 agent 直接向商户付小额，取代靠广告养活的免费网络。把支付和发现绑在一起的工具（如 AgentCash）有机会取代广告。",
   "cav": "作者 Sam Ragsdale 是前 a16z crypto 同事、现 Merit Systems（AgentCash）创始人，文章发在 a16z crypto 渠道，利益相关；a16z crypto 官网列表标 2026-03-20，此处按 Substack 页 2026-03-21。",
   "stance": "稳定币 vs 卡：开放协议加稳定币替代广告模式；需求：被当作趋势判断，没有给出量化证据。",
   "vs": "一致",
   "why": "开放协议加稳定币让 agent 直接付小额，替代广告养的免费网络",
   "rail": "稳定币",
   "region": "美国"
  },
  {
   "t": "Agentic commerce won't kill cards, but it'll open a gap",
   "org": "a16z crypto",
   "au": "Noah Levine",
   "date": "2026-03-04",
   "url": "https://a16zfrontier.com/posts/article/agentic-commerce-wont-kill-cards",
   "kind": "赛道论点",
   "seg": [
    "all",
    "s0",
    "m7"
   ],
   "co": [],
   "gist": "反驳「稳定币会替代 Visa/万事达」：卡提供授信、预授权、反欺诈和拒付保护，稳定币现在都没有，所以大部分交易仍是卡赢。真正的机会在卡网络和处理商承保不了的商户（无网站、无主体的 AI 原生卖家），稳定币基础设施会在这个缺口里长大。",
   "cav": "页面日期格式 03.04.26，按 a16z 同站其他文章的月.日.年格式读作 2026-03-04；该文旧域名 a16zcrypto.com 已 301 到 a16zfrontier.com。",
   "stance": "稳定币 vs 卡：卡赢主流，稳定币赢卡覆盖不到的商户；需求：真实但先在缺口商户出现。",
   "vs": "部分",
   "why": "卡的授信、拒付、反欺诈稳定币都没有，主流交易卡赢；新层只长在卡覆盖不到的商户",
   "rail": "并存",
   "top": 1,
   "region": "美国"
  },
  {
   "t": "Agents will pay like locals, not tourists",
   "org": "a16z crypto",
   "au": "Sam Broner",
   "date": "2026-02-28",
   "url": "https://a16zfrontier.substack.com/p/agents-arent-tourists",
   "kind": "赛道论点",
   "seg": [
    "all",
    "s0"
   ],
   "co": [],
   "gist": "Agent 更像企业而不是游客：会靠预先谈好的供应商关系、信用和账期交易，而不是每次现场刷卡。卡网络管不好很小额和很大额，可编程、低成本的稳定币最先在这些新场景落地，创业者该补上对账单、仲裁、批量审批这些缺口。",
   "cav": "同文在 a16z crypto 官网的标题是《Tourists in the bazaar…》，日期 2026-02-19；这里是 a16z Substack 转载页读到的标题和日期（2026-02-28）。",
   "stance": "稳定币 vs 卡：B2B 与长尾场景稳定币先到；需求：靠大平台的谈判能力，规模效应明显。",
   "vs": "部分",
   "why": "agent 像企业一样靠账期和供应商关系交易，新轨道只先补很小额和很大额",
   "rail": "并存",
   "region": "美国"
  },
  {
   "t": "Powering the Agentic Economy: Our Seed Investment in Sapiom",
   "org": "Accel",
   "au": "Amit Kumar、Pierce Daly",
   "date": "2026-02-06",
   "url": "https://www.accel.com/news/powering-the-agentic-economy-our-seed-investment-in-sapiom",
   "kind": "投资备忘",
   "seg": [
    "n8",
    "w1",
    "p3"
   ],
   "co": [
    "Sapiom"
   ],
   "gist": "软件不再只干活还会花钱：买算力、调用按量计费的 API、买数据，形成连续、自动、规模前所未有的新类别支出；现有金融系统默认有人，需要一层有治理、可编程的支付层，Sapiom 用一次集成给 agent 带策略控制的付费服务访问。",
   "cav": "领投种子轮，媒体报道约 1500 万美元；页面日期 2026-02-06（搜索摘要为 02-05）。",
   "stance": "需求：开发者/agent 买 API 与算力是第一批场景。",
   "vs": "一致",
   "why": "软件花钱买 API、算力、数据是新类别支出，要有治理的支付层",
   "rail": "不涉及",
   "region": "美国"
  },
  {
   "t": "The Agentic Payments Map",
   "org": "Fintech Brainfood",
   "au": "Simon Taylor",
   "date": "2026-02-01",
   "url": "https://www.fintechbrainfood.com/p/the-agentic-payments-map",
   "kind": "市场地图",
   "seg": [
    "all",
    "p3",
    "g4"
   ],
   "co": [],
   "gist": "协议这么多是因为它们解决不同的问题、处在不同的层：信任与身份、商务协议、支付轨道，需要跨层一起用，而不是选一个。先有信任层（知道对面是谁），agent 才能付款；标准还不成熟，可能会淘汰一批，商户先看商务协议。",
   "cav": "Simon Taylor 是 Fintech Brainfood 作者、投资人兼评论员，不是基金官方发布；页面日期读到 2026-02-01，搜索摘要有标 01-31。",
   "stance": "稳定币 vs 卡：不站队，各层并存；需求：认为 agent 原生支付会长成交易的重要份额，但标准还早。",
   "vs": "一致",
   "why": "信任与身份、商务协议、支付轨道要跨层一起用，先有信任层才能付款",
   "rail": "并存",
   "top": 1,
   "region": "欧洲"
  },
  {
   "t": "Paid Raises $21M Seed Round to Build the Business Engine for the Agent Economy",
   "org": "EQT Ventures",
   "au": "",
   "date": "2025-09-29",
   "url": "https://stories.eqtventures.com/articles/paid-raises-21m-seed-round-to-build-the-business-engine-for-the-agent-economy",
   "kind": "投资备忘",
   "seg": [
    "b9"
   ],
   "co": [
    "Paid"
   ],
   "gist": "agent 接管原本由人团队做的事，按席位的 SaaS 定价会失灵，企业需要按结果计费的体系；EQT 引用的客户案例（Artisan、Boon、IFS）称采用 Paid 后收入增长 20 到 40%，作为继续加注的理由。",
   "cav": "20–40% 是文中转述客户成绩，属公司自报；页面元数据日期与正文日期冲突，取正文 2025-09-29（Lightspeed 领投种子轮）。",
   "stance": "需求：用客户收入增长佐证。",
   "vs": "一致",
   "why": "按席位定价失灵，要按结果计费",
   "rail": "不涉及",
   "region": "欧洲"
  },
  {
   "t": "The AI Agent Economy Has a $19 Trillion Problem: Our Investment in Paid",
   "org": "Lightspeed Venture Partners",
   "au": "Alexander Schmitt",
   "date": "2025-09-28",
   "url": "https://lsvp.com/stories/the-ai-agent-economy-has-a-19-trillion-problem-our-investment-in-paid/",
   "kind": "投资备忘",
   "seg": [
    "b9"
   ],
   "co": [
    "Paid"
   ],
   "gist": "Agent 构建者没法衡量和定价自己 agent 干的活，导致价值流失；按席位的 SaaS 定价不适合连续运行、结果差异大的 agent。Paid 提供按客户定价、按结果计费和成本追踪，Lightspeed 因此领投，并看好创始人 Manny Medina。",
   "cav": "领投 2100 万美元种子轮；标题中的「19 万亿」为文中估算，未逐项核实。",
   "stance": "需求：结果计费是 agent 公司的刚需。",
   "vs": "一致",
   "why": "agent 干的活没法计量和定价，要新的计费层",
   "rail": "不涉及",
   "region": "美国"
  },
  {
   "t": "Our Investment in Kite",
   "org": "General Catalyst",
   "au": "Marc Bhargava",
   "date": "2025-09-02",
   "url": "https://www.generalcatalyst.com/stories/our-investment-in-kite",
   "kind": "投资备忘",
   "seg": [
    "w1",
    "g4",
    "p3"
   ],
   "co": [
    "Kite AI"
   ],
   "gist": "AI agent 将承担网络上越来越多的活动，互联网底层要为 agent 加上身份、可编程治理和支付；GC 领投种子轮并加注 1800 万美元 A 轮，看好 Kite 的信任层，及其 Kite AIR 已与 Shopify、PayPal 集成。",
   "cav": "同轮 PayPal Ventures 也发了文章，但页面打不开未收。",
   "stance": "需求：用已有的 Shopify、PayPal 集成佐证。",
   "vs": "一致",
   "why": "互联网底层要为 agent 加身份、可编程治理和支付",
   "rail": "不涉及",
   "region": "美国"
  },
  {
   "t": "Investment: Skyfire, Powering Agentic Commerce",
   "org": "Blitzscaling Ventures",
   "au": "Jeremiah Owyang",
   "date": "2025-09-02",
   "url": "https://www.blitzscalingvc.com/post/we-invested-in-skyfire-powering-agentic-commerce",
   "kind": "投资备忘",
   "seg": [
    "w1",
    "g4",
    "n8"
   ],
   "co": [
    "Skyfire"
   ],
   "gist": "近期大量金融交易将由经过验证的 AI agent 代为完成；Skyfire 把「agent 对应一个负责任的真人」的身份验证、嵌入式钱包和 USDC 支付合成安全的 agent 商业基础设施。团队此前做的 Jibe 被 Google 收购，后来在 Ripple 的经历也是信心来源。",
   "cav": "小型早期基金的博客；日期读到 2025-09-02。",
   "stance": "稳定币 vs 卡：USDC 作为结算资产；需求：判断为趋势。",
   "vs": "一致",
   "why": "把 agent 对应到负责任的真人，再配钱包和 USDC 支付",
   "rail": "稳定币",
   "region": "美国"
  },
  {
   "t": "Investing in Catena Labs",
   "org": "a16z crypto",
   "au": "Chris Dixon、Elizabeth Harkavy",
   "date": "2025-05-20",
   "url": "https://a16zfrontier.com/posts/article/investing-in-catena-labs/",
   "kind": "投资备忘",
   "seg": [
    "w1",
    "g4",
    "s0"
   ],
   "co": [
    "Catena"
   ],
   "gist": "Agent 正成为自主经济主体，但现有支付默认每笔有人在场，对自动化活动慢、贵、还常被当成欺诈；Catena 做面向 agent 的新金融栈，让可信软件 agent 用稳定币全球收付，结算快、费用低、内置身份与合规。",
   "cav": "同一篇在 Agent Infra 图谱的 vc.js / from-infra.json 里已有（链接域名为旧的 a16zcrypto.com，现已 301 到 a16zfrontier.com），已重新打开核对。",
   "stance": "稳定币 vs 卡：押稳定币；需求：看好 AI 内容实时付款、按需买数据服务。",
   "vs": "一致",
   "why": "现有支付默认每笔有人在场，自动化交易慢、贵、常被当欺诈",
   "rail": "稳定币",
   "region": "美国"
  },
  {
   "t": "Why Agents Need a New Payment Stack — and Why Nekuda is Building It",
   "org": "Madrona",
   "au": "Jon Turow",
   "date": "2025-05-14",
   "url": "https://www.madrona.com/agents-need-a-payment-stack-nekuda-is-building-it/",
   "kind": "投资备忘",
   "seg": [
    "g4",
    "k6"
   ],
   "co": [
    "Nekuda"
   ],
   "gist": "Madrona 花了约 18 个月投 agent 创始人，认为 agent 会改变钱在软件里的流动方式；Nekuda 给 agent 提供网络层面强制执行的认证和限额，是缺失的信任层。Amex Ventures、Visa Ventures 的跟投和 Visa Intelligent Commerce 的首发合作让 Madrona 相信需求真实。",
   "cav": "领投 500 万美元；Visa、Amex 的跟投让立场偏卡网络生态。",
   "stance": "稳定币 vs 卡：站卡网络加 agent 凭证；需求：有 Visa 合作佐证。",
   "vs": "一致",
   "why": "缺的是网络层面强制执行的认证和限额这层信任",
   "rail": "卡",
   "region": "美国"
  },
  {
   "t": "Rewriting the Rules of AI Monetization: Why We're Backing Paid's €10M Pre-Seed Round",
   "org": "EQT Ventures",
   "au": "Doreen Huber",
   "date": "2025-03-25",
   "url": "https://stories.eqtventures.com/articles/rewriting-the-rules-of-ai-monetization-why-we-re-backing-paid-s-10m-pre-seed-round",
   "kind": "投资备忘",
   "seg": [
    "b9"
   ],
   "co": [
    "Paid"
   ],
   "gist": "AI 原生业务需要和 SaaS 不同的变现模式：按结果而非按席位或用量定价；EQT 把 Paid 看作 AI 经济的金融基础设施，信心来自创始人 Manny Medina 把 Outreach 做成市场领导者的经历。",
   "cav": "页面元数据日期与署名日期冲突，这里取署名日期 2025-03-25；该轮由 EQT 与 Sequoia 共同领投 1000 万欧元。",
   "stance": "需求：结果定价刚性需求。",
   "vs": "一致",
   "why": "AI 原生业务要按结果定价，不是按席位或用量",
   "rail": "不涉及",
   "region": "欧洲"
  }
 ]
}