Levchin Sees AI Payment Agents Displacing Credit Card Terminals

莱夫琴认为AI支付智能体将取代信用卡终端 Affirm首席执行官Max Levchin表示，人工智能将通过动态路由支付与融资渠道来重塑结账环节，而非自主替消费者选择购买商品。 Max Levchin与Alex Rampell回顾了塑造现代数字支付的结构性力量，解释了为何信用卡交互界面极难被取代，以及支付领域的微小细分赛道为何能演变为庞大市场。他们梳理了Affirm早期的商业探索历程——从解决移动端购物车弃购率（“睡衣问题”），到首创基于身份的核保模式，再到将分期选项前置以提升商户转化率。最后，讨论回顾了PayPal黑帮独特的文化基因，并将确定性较低的自主“智能体购物”与具备高度确定性的“智能体支付”进行了对比。 尽管支付是全球最大的市场之一，但已有 60 年历史的信用卡仍被视为极难颠覆的交互界面；Apple Pay 与 Google Pay 的普及，很大程度上归功于强制更换 EMV 芯片终端时意外普及了非接触式刷卡硬件。 (1:33) Affirm 的创立初衷是为了解决消费者卧床浏览网购却懒得起身拿实体卡的“睡衣问题”，通过将社交与数字身份转化为承保依据，在用户购买意图最强的节点提供极低摩擦的即时信贷。 (17:15) Affirm 将分期选项直接前置到商品展示页带来了 30% 的转化率爆发，并依托高毛利 DTC 床垫品牌的爆发确立了无隐藏罚息的真实 0% APR 模式，最终跑通了具备“负获客成本”（Negative CAC）的商业壁垒。 (33:00) 对于 AI 前景，Max Levchin 极度看好“代理支付”（Agentic Payments）替代传统卡片界面并动态优化结算，但对完全接管选品的“代理购物”（Agentic Shopping）持怀疑态度，因消费者依然渴望保留审美与发现商品的自主权。 (52:44)

Max Levchin and Alex Rampell argue that while the physical credit card remains the most resilient user interface in payments, autonomous AI payment agents—rather than shopping agents—could finally replace it by automating checkout routing and financing. The Resilience of the Credit Card Interface (0:00) The credit card interface has dominated commerce for over six decades because low-ticket, high-frequency purchases—measured by the canonical "cup of coffee" test—demand instantaneous, frictionless execution. Visa and Mastercard mandate a strict 2.5-second authorization window, leaving virtually no tolerance for clunky authentication. Levchin and Rampell point out that physical payment novelties, from fuel-pump RFID wands to Amazon's palm-scanning at Whole Foods, consistently failed to achieve critical mass because they offered only marginal speed improvements over tapping a card or phone. Changing consumer payment habits at scale requires systemic infrastructure shifts rather than novelty. Apple Pay and Google Pay gained ubiquity primarily because the EMV liability shift penalized merchants for magnetic-stripe counterfeit fraud, forcing businesses to install chip terminals that happened to include contactless tap hardware. Combined with smartphone secure enclaves and hygiene preferences during COVID-19, tap-to-pay became standard, demonstrating that payment habits change only when merchant infrastructure, risk policy, and consumer hardware align. Solving Checkout Friction with Up-Funnel Installments (29:57) Affirm originated to solve the "pajama problem"—the high rate of mobile cart abandonment caused by consumers in bed refusing to get up and retrieve a physical credit card. Levchin and Rampell initially sought to revive the 19th-century general store tab by underwriting consumers based on social identity signals rather than traditional FICO scores. However, early trials offering point-of-sale loans at checkout—such as with 1-800-Flowers—merely cannibalized standard card transactions while imposing a steep 7% merchant discount rate. Product-market fit materialized when online cosmetics merchant Beautylish moved installment offers up-funnel directly onto product pages, driving an immediate 30% lift in checkout conversion. This transformed the product from an identity-checkout workaround into a budget-expansion engine. Affirm capitalized on the direct-to-consumer mattress boom (Casper and Purple), where 80% gross margins and seven-year replacement cycles justified merchants absorbing high processing fees to offer 0% APR financing over three years, while explicitly rejecting the deferred-interest penalties and hidden fees common in traditional store cards. Negative Acquisition Costs and Servicing Moats (42:15) By integrating directly into merchant checkouts, Affirm evolved into an upstream demand engine with over 50 million consumers across four countries, acquiring users at a negative customer acquisition cost. While platforms like Netflix or Zynga historically resisted third-party payment intermediaries owning customer interactions, retail merchants actively delegate financing to Affirm. Brands selling premium goods deliberately offload loan servicing to avoid damaging their brand affinity with dunning calls or debt collection. This dynamic grants Affirm a persistent, trusted relationship spanning 12 to 39 billing cycles on multi-year loans. Levchin explains that underwriting multi-year loans requires proprietary machine-learning models rather than simple FICO lookups, establishing a structural moat. Each recurring repayment touchpoint doubles as an engagement opportunity to cross-sell financial services, merging payments with customer acquisition. Agentic Payments Over Autonomous Shopping (52:44) Looking toward future interfaces, Levchin distinguishes between "agentic shopping" and "agentic payments." He remains skeptical of AI agents autonomously selecting discretionary goods because consumers genuinely enjoy the identity and curation of browsing style or hobby gear. Conversely, both Levchin and Rampell see immense upside in agentic payments, where software completely removes transaction friction by dynamically selecting optimal financing terms and payment routing. While Rampell notes that consumers will readily delegate price-comparison tasks for commoditized products with standard SKUs, Levchin cautions that delivery reliability, return friction, and merchant reputational risk still impede full automation. However, high-frequency services like Instacart have already conditioned shoppers to accept delegated substitutions for routine groceries, indicating that AI agents will first command automated replenishment before conquering broader consumer commerce. While AI agents are well-positioned to automate financing decisions and streamline checkout routing, the full displacement of legacy card interfaces will depend on whether automated systems can reliably manage merchant trust, delivery logistics, and returns.

Alex Rampell与Max Levchin认为，实体信用卡凭借极快速度与极简交互统治了支付界面长达60年，但随着AI代理在支付路由与动态信贷承保上的突破，自动化支付有望首次真正替代物理卡片界面。 信用卡交互的长期统治与微交易摩擦 (0:00) 支付是全球规模最大的市场，但处理体量并不等同于高利润率。高利润往往集中在快餐等小额高频交易，而非费率极低的大额B2B转账。信用卡清算网络（Visa与Mastercard）对网络、发卡行、商户及收单机构的授权响应设定了严格的2.5秒限制，全流程超时上限仅为10秒。线下POS系统的极端时间约束，使星巴克等商户选择通过50美元预充值来规避单笔咖啡刷卡产生的高昂网络费。Rampell指出，改变消费者支付习惯极度困难，Apple Pay与Google Pay的崛起并非单纯源于用户意愿，而是受EMV芯片责任转移政策驱动：商户为避免承担伪造磁条卡的全额盗刷损失被迫更新终端，意外普及了芯片内自带的非接功能，再结合手机本地安全隔离区和疫情无接触需求才完成习惯迁移。 Levchin强调，支付创新具有赢家通吃或临界规模特征，微小的便利性提升很难撬动昂贵的终端替换成本，亚马逊在全食超市推广的掌纹识别等硬件最终多沦为过渡噱头。在加密货币领域，尽管比特币作为价值存储已获巨大成功且稳定币实用性显现，但加密体系始终无法通过“买咖啡”这一基准测试。在小额高频场景中，极致的低摩擦和便利性压倒一切底层技术，任何需要繁琐密码或等待确认的体验都会促使消费者迅速退回传统的刷卡或零钱支付。 Affirm演变：从睡衣痛点到预算扩张 (17:15) Affirm的创立源于2009至2012年间对移动电商“睡衣问题”的观察：用户在床上浏览手机时，往往因懒得起身取信用卡而放弃购物车。创始人借鉴19世纪杂货铺依据熟人声誉提供赊购的逻辑，试图将Facebook好友数和照片作为社交信用凭据，在结账瞬间完成无抵押授信。早期的1-800-Flowers合作测试因7%的高额商户费率和转化率不足遭遇质疑，商户指责其仅仅蚕食了既有的信用卡交易额。 真正的转折点发生在美妆平台Beautylish将分期提示前置展示在商品详情页，直接推动结账转化率提升30%。这使Affirm明确了自身本质是促进商户销售的预算扩张工具。伴随Casper和Purple等毛利率高达80%的DTC床垫品牌兴起，由于消费者平均每7年才更换一次床垫，品牌商极其渴望通过承担高额手续费提供3年期0% APR分期来锁定长周期客户。Levchin在此确立了彻底消除隐藏费用和追溯性罚息的透明信贷承诺，与传统百货商店带有滞纳陷阱的“虚假0% APR”延期利息模式划清了界限。 负获客成本与信贷关系构建的壁垒 (42:15) 不同于多数将大量资金输送给Google与Meta作为广告费的消费类创业公司，Affirm直接嵌入商户结账流程，在获取数千万高意向消费者的同时向商户收取折扣费，实现了罕见的“负获客成本”。Rampell指出，传统B2B2C模式下平台方往往极力防范第三方介入客户关系，但在大额分期场景中，床垫等品牌商既希望保证顾客体验，又极力避免因向逾期用户催收而损害品牌形象，因而主动将长达数年的信贷履约与服务沟通委托给Affirm。 这种分期履约在12至39个月的周期内为Affirm创造了持续、可信的用户互动触点，使其能够低成本交叉销售其他金融服务。Levchin指出，长达3.5年的长周期贷款无法仅靠简单的FICO评分或社交信号支撑，必须依赖严格的机器学习风控算法，这构成了防范竞争对手的防御护城河。目前Affirm已积累超过5000万交易用户，其支付网络也逐步演变为能够为商户精准锁定购买力人群的需求生成平台。 代理人支付：重构交易界面的真正契机 (52:44) 展望金融科技未来，Levchin对全自主的“代理人购物”持怀疑态度，但对“代理人支付”极为看好。他认为，消费者在选购服饰、骑行配件等生活品类时，核心诉求在于享受探索过程与表达个性品味；然而在确认购买后，掏卡、录入信息与选择信贷等结账操作纯属机械摩擦。虽然实体塑料卡作为支付界面已垄断长达60年，但能够动态匹配最优分期方案和实时资金路由的AI代理，有望真正替代物理卡片。 Rampell指出，在标准SKU或条形码明确的标准化商品上，追求时间效率的消费者很乐意将比价与下单托管给AI。Levchin补充说明，尽管非知名网站的履约可靠性与退货风险仍制约着AI代购的全面铺开，但Instacart等代客买菜服务的普及已经证明，大众愿意在常规消费中接受由第三方代为挑选替代品的决策逻辑。随着AI逐步掌握逆向物流与售后处理，标准化领域的代理人交易将加速渗透。 虽然AI代理在消除结算摩擦和优化支付路由方面展现出颠覆传统信用卡的巨大潜力，但在非标消费领域的审美探索、长尾商户的履约信用风险以及复杂的退换货机制，仍是自动化交易全面普及前有待解决的现实制约。

## Background 本期内容由投资机构 a16z 于 2026 年 9 月 3 日在 YouTube 频道发布。节目由主持人 Erik Torenberg 邀请 a16z 普通合伙人 Alex Rampell 与 Affirm 联合创始人兼 CEO Max Levchin 共同对谈，深入探讨了过去 25 年金融科技的演进历程、Affirm 从零创立的创想迷宫以及 AI Agent 对未来支付与商业重塑的前景。 ## TL;DR Max Levchin 与 Alex Rampell 回顾了数字支付 25 年的底层逻辑与历史巧合，解析了 60 年未变的信用卡规则为何坚如磐石，以及为何即便微小的支付分支也能成长为千亿美元级市场。他们完整复盘了 Affirm 如何从最初解决睡袍痛点与社交身份信用摸索，最终转型为依托漏斗上游分期来扩张用户预算、凭借真正的 0% 利率和负获客成本构筑护城河的历程。此外，Max 分享了关于 PayPal 黑帮的真实切片，并指出相比于由 AI 代替人类挑选非标商品的代理购物，AI 代理更具确定性的颠覆方向在于重构塑料信用卡长达数十年的底层支付与路由界面。 ## Atmosphere & style 整场对谈兼具老友叙旧的幽默松弛与顶级从业者的硬核洞察，两位嘉宾不仅时常拿早年的俄语邮件、破产派对和卖床垫的往事互相打趣，还毫无保留地解构了埃隆·马斯克与彼得·蒂尔等硅谷巨擘在早期创业高压下卸下光环、狼狈脆弱的真实面貌。叙事节奏层层递进，既有对商业数字、转化率与风控底层架构的冷静剖析，也有对创业穿越荒漠至暗时刻的真挚感触，带给观众一种沉浸式旁听行业教父级人物复盘历史与推演未来的通透感。 ## Bulletin - [00:00](https://www.youtube.com/watch?v=J3pegsM5drk&t=0s) Max Levchin 与 Alex Rampell 指出，支付赛道不存在小于 1000 亿美元的利基市场，但其商业逻辑呈现显著倒挂：超大额交易费率极薄，而小额高频日常消费的便利性溢价才是核心利润来源。 - [33:00](https://www.youtube.com/watch?v=J3pegsM5drk&t=1980s) Affirm 商业模式的根本转折发生于从漏斗下游结账转移到商品展示上游，将业务本质从解决忘记带卡的“睡衣难题”重新定义为提供预算扩容的消费金融，直接带动商户转化率飙升 30% 以上。 - [44:05](https://www.youtube.com/watch?v=J3pegsM5drk&t=2645s) 通过让高毛利商户主动承担服务费以拉动销售，Affirm 构建了极其罕见的“负获客成本”（Negative CAC）壁垒，不仅无需向广告巨头采买流量，反而借由 5000 万消费者建立起多达 39 次持续互动的金融契约关系。 - [52:44](https://www.youtube.com/watch?v=J3pegsM5drk&t=3164s) 在 AI 浪潮下，Levchin 看淡“AI 代替用户挑选主观商品”（Agentic Shopping），但强烈看好“代理支付”（Agentic Payments）颠覆沿用 60 年的信用卡底层规则与资金路由机制。 ## Key takeaways - [10:00](https://www.youtube.com/watch?v=J3pegsM5drk&t=600s) 信用卡塑料卡片被公认为有史以来最卓越的单一用户界面（UI），历经生物识别、感应棒与加密货币等多次尝试均难以被撼动。 - [01:01](https://www.youtube.com/watch?v=J3pegsM5drk&t=61s) 移动非接触支付（Tap）的普及是一系列巧合的产物，主要由 EMV 芯片责任转移迫使商户被动升级终端、插卡体验迟缓以及疫情对卫生的需求共同催化而成。 - [02:13](https://www.youtube.com/watch?v=J3pegsM5drk&t=133s) 自 Dee Hock 创立 Visa 以来，卡组织强行设定的 2.5 秒交易响应上限已固化 60 年未变，Apple Pay 凭借芯片安全区“时间平移”预处理才实现了体验破局。 - [12:36](https://www.youtube.com/watch?v=J3pegsM5drk&t=756s) 加密货币在价值存储和稳定币结算上展现了价值，但因 UI 交互摩擦与小额日常支付对便利性的极致要求，始终无法通过“买咖啡测试”。 - [05:54](https://www.youtube.com/watch?v=J3pegsM5drk&t=354s) B2B 借贷与供应链保理中，小供应商被迫承受高达 1000 个基点的资金利差，但 B2B 融资因参与者高度理性，实际利润空间远低于具备便利性溢价的消费金融。 - [20:01](https://www.youtube.com/watch?v=J3pegsM5drk&t=1201s) Affirm 的最初构想是复刻 19 世纪小镇杂货店的赊账文化（Pay with your identity），试图借助 Facebook 社交图谱和反欺诈机器学习模型在移动端解决躺在床上不愿拿卡的“睡衣难题”。 - [33:00](https://www.youtube.com/watch?v=J3pegsM5drk&t=1980s) 在结账漏斗底部遭遇转化率低与商户抵触的高昂费率后，美妆电商 Beautylish 将分期提示透传至商品详情页上游，触发转化率暴增 30%，使团队确立了“预算扩容”的真正产品市场契合点（PMF）。 - [35:54](https://www.youtube.com/watch?v=J3pegsM5drk&t=2154s) 以 DTC 盒装床垫为代表的高毛利（80%+）品类因 7 年长换新周期引发激烈获客战，驱动商家自愿承担高额商户扣率（MDR）来补贴 0% APR 免息分期。 - [40:37](https://www.youtube.com/watch?v=J3pegsM5drk&t=2437s) Affirm 严厉剔除传统信用卡行业依靠小字星号、递延利息（Deferred Interest）和逾期滞纳金盈利的“伪 0%”陷阱，以绝对透明的前端总价锁定赢得市场信任。 - [44:46](https://www.youtube.com/watch?v=J3pegsM5drk&t=2686s) Affirm 摆脱了传统消费类创业公司向 Google 和 Meta 纳贡的困境，通过向渴望拉升 GMV 的商家收费实现“负获客成本”（Negative CAC），并在长达 42 个月的贷款周期中获得 12 至 39 次持续互动的运营机会。 - [48:18](https://www.youtube.com/watch?v=J3pegsM5drk&t=2898s) “PayPal 黑帮”的孕育得益于刻意筛选具备创业野心人才的机制，以及早期团队在极度高压的白板争吵中目睹彼此作为普通人的脆弱与挣扎，从而彻底消除了对宏大目标的畏惧感。 - [55:44](https://www.youtube.com/watch?v=J3pegsM5drk&t=3344s) 在主观性、非标性消费中，用户视挑选为核心乐趣，且长尾电商网站存在不可忽视的履约与退货信任鸿沟，导致纯粹由 AI 替客比价购物的普及速度将远低于预期。 - [57:48](https://www.youtube.com/watch?v=J3pegsM5drk&t=3468s) Instacart 等代客买菜平台证明全流程代理消费已在现实中发生；一旦解决退货、真伪鉴别与物流履约等现实阻力，代理化支付与商务将真正重塑零售业态。 ## Detailed walkthrough ### [00:00](https://www.youtube.com/watch?v=J3pegsM5drk&t=0s) 开篇引言：信用卡界面的不可动摇与百亿美元级别的支付微细市场 信用卡支付界面被公认为“有史以来创造出的最优秀的单一用户界面（UI）”，而整个支付领域也毫无争议地构成了全球最庞大的市场。在支付赛道中，根本不存在规模小于 1000 亿美元的所谓“利基市场”（niche）。然而，这一领域也存在着奇特的倒挂现象：资金规模极度庞大的交易，费率和利润空间反而极度微薄；那些真正蕴藏着巨额交易量与丰厚收入机会的，往往是小额高频的日常消费。随着支付金额逐渐变小，便利性压倒了一切考量。 如今，随着人工智能的发展，这个数十年来坚不可摧的信用卡交互界面或许终于迎来了被“重新谈判”的契机——尽管核心的瓶颈在于，用户目前尚未完全建立起对 AI Agent 的信任，不敢放手让 Agent 做出与人类同样出色的决策。Max Levchin 回忆起当年创立 PayPal 时的一段插曲：他曾在某次密码学学术会议上公开展示一种全新的数字支付方案，结果当场“被台下嘘下了台”（literally booed off stage），原因就在于密码学圈子执着于匿名性，而他的方案完全抛弃了匿名特征——正是这种对匿名的反向思考，奠定了 PayPal 颠覆支付体系的基石。 --- ### [01:01](https://www.youtube.com/watch?v=J3pegsM5drk&t=61s) 25年金融科技回望：Apple Pay、EMV芯片与消费者行为的意外重塑 回溯两人在金融科技领域超过 25 年的探索历程，主持人 Erik Torenberg 询问究竟有哪些现象超出了他们当年的预料。Alex Rampell 首先指出，最令他感到意外的是 Apple Pay 和 Google Pay 竟然完成了如此深度的市场渗透。从历史经验来看，扭转大众的消费习惯几乎是不可能完成的任务，而移动非接触式支付的普及，实际上源于一系列戏剧性的“历史巧合”的叠加： 这一转折的核心驱动力，是早年由 Visa、Mastercard 以及 Europay 共同推进的 EMV 芯片标准与“商户欺诈责任转移”（Merchant Liability Shift）机制。传统的磁条卡极易被复制，在老旧规则下，消费者如果在百思买（Best Buy）刷磁条卡买了一台电视，回家后完全可以谎称自己从未购买，百思买若无法自证便只能咽下全部损失。为了规避这种被恶意薅走电视的赔付责任，全美商户被迫大规模更换刷卡终端。 随后出现的局面极具戏剧性： ```mermaid flowchart TD accTitle: 移动非接触支付普及的多因素协同演进图 accDescr: 展示EMV芯片责任转移、终端升级与智能手机及疫情如何共同催化非接触支付的爆发 subgraph 基础设施被动改造 A["传统磁条卡易被复制欺诈"] --> B["EMV责任转移机制实施"] B --> C["全美商户被动更换POS终端"] C --> D["新终端普遍默认附带NFC非接触式感应线圈"] end subgraph 行为催化与终端就绪 E["早期用户体验槽点<br/>插卡芯片写入缓慢（提示请勿拔卡）"] F["智能手机与移动终端全面普及"] G["COVID 疫情爆发驱动非接触卫生需求"] end D --> H["非接触式 Tap 支付瞬间爆发并成为主流"] E --> H F --> H G --> H ``` 当这些新终端就位时，插芯片往往伴随着冗长的写入提示（“正在处理，请勿拔出卡片”），而机器上原本闲置的小小感应区，恰好为挥卡与手机轻触（Tap）铺平了道路。加上 COVID 疫情催化了对非接触交互的需求，以及智能手机的全面普及，这三股力量交织在一起，最终不可逆地重塑了全球的刷卡支付习惯。 --- ### [02:13](https://www.youtube.com/watch?v=J3pegsM5drk&t=133s) 60年未变的规则底色：Dee Hock时代遗留的2.5秒铁律 面对同样的行业演进，Max Levchin 提出了更为犀利的反思。他认为，身处这个行业 30 余年，最不可思议的事情其实是底层规则的顽固不化。自 Dee Hock 在 20 世纪建立 Visa 体系以来，卡组织的底层硬性规则已经沿用了整整 60 年，至今仍未被撼动。 其中最核心的瓶颈，是 Visa 和 Mastercard 强行设定的“2.5 秒响应上限”：从终端发起请求，穿透网络、发卡行、商户以及收单行，整个链路必须在极短时间内完成，整笔交易若在 10 秒内未获响应就会被系统直接取消或重试。 * **线上与线下的差异**：在线上电商结算场景中，商户可以“耍花招”（play games），例如在将卡号提交给卡组织之前先运行后台反欺诈评分与风控模型以规避风险；但在物理世界面对面的结账柜台前，商户与发卡行手里永远只有死板的 2.5 秒，这几乎扼杀了任何复杂的交互与金融创新。 * **Apple Pay 的时间平移创新**：Apple Pay 和 Google Pay 之所以伟大，是因为它们利用手机硬件芯片中的安全隔离区（Secure Enclave），对整条支付链路进行了“时间平移”（timeshifted）。手机在真正向 Visa/Mastercard 广播交易之前，设备本地就已经完成了身份认证与卡片信息解密。 * **规则滞后的缺憾**：令人吃惊的是，卡组织自身至今仍未对 2.5 秒的时限做出演进。例如，将时限放宽至 15 秒，允许系统在交易发生的瞬间撮合多家发卡机构实时竞价，为持卡人争取更好的信贷利率或定制化的权益。 尽管存在种种规则约束，支付行业的惊人之处在于其体量过于庞大，哪怕只是对最微小的细分切面进行解构与微调，最终都能孵化出一个超过 1000 亿美元的独立市场。 --- ### [05:54](https://www.youtube.com/watch?v=J3pegsM5drk&t=354s) 支付的规模悖论与B2B借贷探索：“Pay Me Sooner”与供应链保理 支付市场内部还存在着一个显著的结构性悖论：“交易金额越大，实际产生的费率空间反而越小”。如果一家巨头机构或者像埃隆·马斯克的火星基地需要电汇一笔数万亿美元的资金，这笔交易固然金额惊人，但提供电汇的通道根本不可能收取 2% 的抽成；相反，真正产生丰厚交易收益的往往是快餐店（QSR）这类极度高频的小额交易。星巴克之所以耗费巨资打造星巴克移动支付（Starbucks Pay），正是为了通过“预付充值”模式将多次微额消费合并为单次充值，从而避免向 Visa 和 Mastercard 反复支付高额小额手续费。 围绕这一规律，Max Levchin 与 Alex Rampell 回顾了他们早年的一段往事。Max 曾试图在旧邮件记录中寻找 Affirm 的最初源头——一封写于 2011 年 4 月的往来邮件。当时 PayPal 刚刚收购了针对消费者的信用支付产品 Bill Me Later，Max 写信向 Alex 详细请教其运作机制。两人经过密集的邮件头脑风暴，甚至一度构想将这种赊购模式照搬至企业端（Bill Me Later for Business），但深入测算后却认定这在商业上完全不可行。 Alex 笑着补充道，自己当时甚至为此抢注了 `paymesooner.com` 这一域名。因为企业端真正存在的痛点并不是“支付”，而是“信贷利差与账期压迫”： * **大企业与供应商的不对等博弈**：类似通用电气（GE）这样的大型核心企业，往往会向小微供应商强制要求长达 90 天（Net 90）的付款账期。 * **资金成本倒挂的荒谬困境**：小企业在完成供货后，为了给员工发放工资，不得不拿着对 GE 的应收账款去商业银行贷款。银行往往开出高达 15% 的借款利率；然而与此同时，GE 自身在债券市场上按担保隔夜融资利率加点（SOFR + 10 bps）发债，借贷成本仅为 5%。 * **商业机会的本质**：小微企业承担的信贷风险，实质上是 GE 的信用风险，但却不得不承受 15% 与 5% 之间高达 1000 个基点的利差剥削。这一错配正是供应链金融与传统保理（Factoring）业务长期试图解决的核心命题。 ### [09:01](https://www.youtube.com/watch?v=J3pegsM5drk&t=541s) 应收账款融资与消费级借贷的利润落差 商业世界中许多借贷需求的根源在于账期的不对称：企业必须等待上下游付款，因此常常陷入“因为别人没按时付钱给我，我才不得不去借钱”的循环。尽管通过折价出售应收账款（Receivables Financing / Factoring）在法律定义上不属于传统放贷、不受高利贷法规限制，并且如今市场上也涌现出一批做应付账款和应收账款融资的成熟企业，但在实际盈利空间上，B2B 借贷与消费金融存在着云泥之别。 消费级借贷之所以暴利，是因为其提供了极为直观的“便利性溢价”，且普通消费者对此感知敏锐。反观企业端融资，参与者高度理性，市场上总有竞品愿意压低费率来争夺单子。正如一笔高达数万亿美元的巨额电汇，其转账规模虽然惊人，但提供底层清算管道的机构却赚不到多少钱——费率被压缩到极低，利润空间微薄。 --- ### [10:00](https://www.youtube.com/watch?v=J3pegsM5drk&t=600s) 支付界最难被颠覆的界面：实体卡片与生物识别的困境 被问及“有哪些理念看似理所应当存在、却至今未能实现”时，马克斯·列夫琴（Max Levchin）提到了人们对生物识别支付的漫长执念。流行文化和好莱坞科幻电影总喜欢描绘切下手指、挖出眼球来通过身份验证的桥段，但现实中人们买单时依然在刷塑料卡片芯片。列夫琴感慨道，卡片支付界面堪称“人类有史以来创造的最卓越的单一用户界面（UI）”，其地位极其坚固，历经无数次颠覆尝试却依旧屹立不倒。 早在 PayPal 诞生之前，美孚（Mobil）与万事达卡等机构曾推出过名为 Speedpass 的便携感应棒，车主在加油机前挥一下即可完成扣款。列夫琴当时第一眼看到它，曾断言“这绝对会彻底取代信用卡”，然而它最终销声匿迹。这揭示了支付创新的一条铁律：支付领域存在一个事后显而易见、事前却极难预判的“临界规模（Critical Mass）”。在支付网络中几乎没有平庸的折中结局，要么实现全民普及成为基础设施，要么彻底被时间长河吞噬。感应棒虽然方便，却只比直接从兜里掏出信用卡“快了一点点”，这微弱的提速根本无法克服改变用户习惯的巨大阻力。 亚历克斯·兰佩尔（Alex Rampell）随后提到了亚马逊此前在 Whole Foods 推出的掌纹支付（Amazon One）测试。虽然把手掌悬在机器上方像在“看手相”一样滑稽又新奇，但它实际上并没有比刷卡更快，甚至往往更慢；亚马逊最终也停止了部分网点的掌纹支付服务。事实证明，除了智能手机之外，人类至今未能找到比实体卡片更高效的支付身份认证载体。 --- ### [12:36](https://www.youtube.com/watch?v=J3pegsM5drk&t=756s) 从 DigiCash 倒闭派对到比特币：为什么加密货币成不了“买咖啡”的货币 谈及加密货币是否曾带来类似实现宿命梦想的震撼，列夫琴坦承自己通常不是新兴趋势最早的追随者。回顾当年创立 PayPal 的历程，其初衷原本是想在超低功耗芯片上运行某种底层加密原语。当团队终于在极其微弱的芯片算力上跑通加密算法后，却发现它弱小到只能用来加密极小的数据片段——而小体积、需要瞬间加解密的数据，恰恰就是支付交易凭证，PayPal 这才应运而生。 就在 PayPal 启动前夕，列夫琴曾参加过数字现金鼻祖 DigiCash 的破产清算派对。在斯坦福大学的校园里，气氛一片肃杀凄凉，一群悲伤的密码朋克们在路边倒酒悼念，痛惜数字支付的黄金时代已经破灭；而密码学先驱大卫·乔姆（David Chaum）甚至因天才构想（盲签名）遭遇商业滑铁卢而心碎地漫步在帕洛阿尔托街头。不久后，列夫琴带着 PayPal 最初的设计在密码学学术会议上演讲，结果遭遇全场嘘声——因为密码朋克们认为 PayPal 既不够安全，也完全抛弃了匿名性。但列夫琴认清了一个现实：普通大众根本不在乎绝对匿名，他们只想顺利买一杯咖啡或在网上买到商品。 面对中本聪的比特币白皮书，列夫琴和许多数学、密码学极客一样，由衷赞叹其利用区块链巧妙解决拜占庭将军问题的算法之美，但他从一开始就不相信它能成为主流货币或日常支付工具。历经多年演进，比特币在资产价值存储、数字黄金和大宗商品属性上取得了巨大成功，如今稳定币也展现出明确的商业价值，但没有任何一种代币能够通过“买咖啡测试”这一检验通用支付手段的最高试金石。 ```mermaid flowchart TD accTitle: 支付便利性与交易金额的关系模型 accDescr: 图表展示了随着交易金额从大额降至极小额，用户决策核心从成本安全转移到极简便利，导致加密货币等复杂流程在日常小额交易中溃败。 subgraph LargeAmount["大额交易（如跨国电汇）"] A["高额标的"] --> B["安全、低费率、通道可靠性压倒一切"] B --> C["用户愿意忍受复杂认证与漫长等待"] end subgraph SmallAmount["小额高频（如街角买咖啡）"] D["微小金额"] --> E["便利性与 UI 交互摩擦压倒一切"] E --> F{"支付速度决策"} F -->|"助记词太长 / 确认慢"| G["放弃加密钱包"] F -->|"掏兜 / 碰一下芯片"| H["掏出信用卡或零钱"] end LargeAmount -.->|"金额迅速递减"| SmallAmount ``` - **买咖啡理论的核心本质**：小额高频交易构成了人类支付最庞大的基本盘。当交易涉及数万亿美元时，人们会精打细算、设计复杂的安全通道；但当你在街角面包店买一杯咖啡时，如果加密钱包需要输入冗长复杂的密码短语，你立刻就会选择掏出零钱或刷一下信用卡。交易金额越小，UI 交互的便利度越能压倒一切成本与安全说教。 --- ### [17:15](https://www.youtube.com/watch?v=J3pegsM5drk&t=1035s) Affirm 创想迷宫前奏：TrialPay 与社交游戏的免费代币生意 关于 Affirm 创立初期如何在“创想迷宫（Idea Maze）”中摸索前行，亚历克斯·兰佩尔笑称他和列夫琴必须像面对“囚徒困境”一样各自对供词，看看记忆是否吻合。兰佩尔首先回顾了他在创立 Affirm 之前经营的公司 TrialPay。 当时的背景是 Facebook 平台上的社交网络游戏正值爆发期。玩家们热衷于在《FarmVille》（开心农场）这类荒唐而充满魔性的社交游戏里给虚拟好友“扔羊”、买游戏币或打虚拟扑克，但绝大多数人不愿意自掏腰包真金白银地为这种毫无实际生产力的虚拟道具付费。TrialPay 抓住的商机是：如果用户不需要花钱，而是通过完成商业注册——例如换用 Geico 车险、申办一张信用卡或订阅 Netflix——就能免费换取一笔游戏币，这便为双方创造了巨大的套利空间。对于保险公司 Geico 而言，获得一个潜在客户具有极高的获客价值（LTV）；而对玩家来说，保费在哪家交都无所谓，顺便还能拿到一大堆《FarmVille》金币。 兰佩尔凭借其极具侵略性且略带磨人的销售手腕，迅速拿下了市面上几乎所有的社交游戏厂商，将这种替代支付模式深植其中。然而，唯有一家极其强硬的社交应用巨头迟迟无法攻克——那便是由马克斯·列夫琴创办的 Slide。 ### [18:30](https://www.youtube.com/watch?v=J3pegsM5drk&t=1110s) 俄语邮件、神秘 Google 禁楼与初次结缘 Alex Rampell 回忆起与 Max Levchin 最初的交集始于 2009 年 3 月的第一周。当时 Max 正在经营社交应用公司 Slide，而 Alex 则受邀参加 Allen & Company 会议。Alex 当时其实差点放弃参会，因为他的妻子预产期是 2009 年 3 月 16 日，随时可能分娩（儿子最终于 3 月 28 日出生）。但在会议主办方的极力劝说下——甚至许诺“一旦你妻子要生了，我们立刻派飞机送你飞回去”——Alex 最终动身前往。在会上，两人有过短暂接触，虽然 Alex 没能当场说服 Max 让 Slide 成为自己创办的公司 TrialPay 的客户，但他随后用俄语给 Max 写了一封跟进邮件。 Alex 在高中和大学学过俄语，并在俄罗斯工作和生活过一段时间。这封带有语法小瑕疵的手写俄语邮件引起了 Max 的好奇，Max 当时的反应是：“你这家伙怎么会说俄语？”两人由此成了朋友。后来 Slide 被 Google 收购，TrialPay 始终没能签下这个客户。直到有一天，Max 在 Twitter 上发帖表示自己正在寻找下一个方向、征集点子，Alex 便顺理成章地前去联络。两人相约在 Google 园区一栋“严禁任何外来访客进入”的神秘大楼里喝咖啡。Alex 调侃道，既然现在追诉时效早已过去，终于可以公开 Max 当年漠视公司规定的秘密了。 - **2009 年 3 月 Allen & Company 会议**：Alex 冒着妻子随时临产的风险赴会，与 Max 首次接洽。 - **手写俄语破冰**：Alex 亲自撰写的俄文跟进信让 Max 印象深刻，Alex 开玩笑说如果换作今天用 Gemini 或 ChatGPT 写出完美语法，反而不会有这种真实感。 - **Google “禁区”咖啡会**：Max 离开 Slide 后在 Google 期间，两人在不对外开放的 Google 大楼内秘密碰面探讨新项目。 --- ### [20:01](https://www.youtube.com/watch?v=J3pegsM5drk&t=1201s) 睡衣难题与 19 世纪“杂货店模型”的构想 在 Google 大楼的那次咖啡交谈中，两人开始讨论一个核心痛点：随着智能手机普及，在手机上完成支付依然极其困难。他们将其形象地称为“睡衣难题”（The Pajama Problem）——用户晚上穿着睡衣躺在楼上的床上想买东西，但信用卡却放在楼下的钱包里，由于不想下床去拿卡输卡号，交易就此告吹。与此同时，不仅移动设备无处不在，社交网络也全面爆发。两人萌生了一个念头：能否结合社交图谱重塑信用审核机制，重拾 19 世纪小镇杂货店的赊账模式？ Alex 解释道，在 1800 年代的传统杂货店里，店主熟识每一个居民；即使你兜里没钱、家里也没现钞，店主一句“Max，我认得你，记在账上就好”便完成了授信。但在现代互联网世界中，消费者仅仅是一串 Cookie 和 IP 地址；线下哪怕走进沃尔玛，迎宾员再热情也完全不认识你，根本无法提供这种信任。他们最初的假设是：如果一个人在 Facebook 上有 500 个真实好友，上传了上千张生活照片，且本身并不是为了借钱而刻意申请贷款，那么其违约风险大概率极低。传统信贷机构总是给 800 分 FICO 信用分的优质人群寄送再融资信件，而主动在 Google 上搜索“我缺钱、急需贷款”的人往往是高风险用户。将授信场景无缝嵌入消费时刻，让用户“用个人身份支付”（Pay with your identity），成了两人最初的设想。 ```mermaid flowchart TD accTitle: 杂货店模式与现代身份支付对比 accDescr: 对比19世纪熟人杂货店授信与基于移动端及社交图谱的身份支付流程 subgraph Historical ["1800 年代小镇杂货店"] H1["顾客进店但身无分文"] --> H2["店主凭借人际熟识度识别身份"] H2 --> H3["记在账上（Put it on my tab）"] end subgraph ModernProblem ["现代电商痛点：睡衣难题"] M1["用户卧床刷手机想购物"] --> M2["信用卡在楼下 / 流程繁琐"] M2 --> M3["放弃下单，转化率流失"] end subgraph ProposedSolution ["最初的 Affirm 构想：身份支付"] P1["社交图谱验证（Facebook 好友/照片）"] --> P2["被动场景授信（非主动借贷搜寻）"] P2 --> P3["免卡直接结账，解决睡衣难题"] end ``` --- ### [22:08](https://www.youtube.com/watch?v=J3pegsM5drk&t=1328s) 妻子的忠告与“社交版 Bill Me Later”的雏形 Max 坦言，在 Google 的那一年是他极为迷茫和痛苦的反思期。尽管 Slide 运营了 5 年多并在财务上让很多人获得了丰厚回报，但他打心底里觉得自己根本不属于社交媒体领域，那段经历从未真正满足他的创业野心。在他陷入沉思、纠结下一步该做什么时，他的妻子对他说了一番至关重要的话：“你这辈子工作最拼命、但内心最快乐的日子，就是当年在 PayPal 做反欺诈系统的时候。那时虽然累到近乎虚脱，但你眼里有光。”即便 Max 当年曾发誓“再也不做金融服务，再也不碰支付”，妻子依然鼓励他放下誓言，再给金融科技一次机会。这番谈话彻底动摇了他，使他不再排斥涉足近支付（near-payments）领域。 随后，两人开始探讨一种“社交版 Bill Me Later”（Social BML）模式，并发现全球各地都有类似小镇杂货店的信用文化。例如在以色列的社区小超市，顾客如果忘带钱包，只需向收银员亮一下手里的购物袋，说一声“下次再付”，收银员就会记在心里；而在日本，曾经普遍流行顾客将名片留给店主，店主将消费金额写在名片背面作为账单挂账；美国文化里也有“记我账上”（Put it on my tab）的习俗。Max 当时还接触了一家试图模仿中国社会信用分建立新型信用评分机制的初创公司。他们看到电视广告时代，观众被电视内容吸引却无法立即购买，如果在移动端能消除支付壁垒，将冲动消费的距离缩短，就能极大提升商家转化率。 - **反欺诈情怀觉醒**：在妻子的点拨下，Max 意识到自己最擅长且最热爱的是利用数据与机器学习解决反欺诈和风险控制问题。 - **跨文化信用现象**：以色列的“下次给”、日本的“名片记账”以及美国的“Tab”挂账文化，证明了基于身份与信任的递延支付具有强大的通用心理基础。 - **消除转化摩擦**：将交易时间窗口压缩，解决消费者“想买却懒得动身拿凭证”的惰性，直接释放被摩擦力压抑的市场需求。 --- ### [25:15](https://www.youtube.com/watch?v=J3pegsM5drk&t=1515s) 动机的分歧与 Affirm 创始团队的组建 尽管两人对愿景充满共鸣，但各自兴奋的侧重点却截然不同。Alex 作为深耕支付领域的创业者，核心动力在于“搞定交易、走通流程、尽快跑去敲开商家的门”；而 Max 刚刚从社交网络的海量机器学习中抽身，满脑子想的是在金融领域大显身手。Max 并不热衷于从头做借贷或传统支付通道，他的理想是“打造一个极其出色的新型信用评分模型，只要建立起顶级的风控算法，放贷和支付通道完全可以交给别人承接”。当时 Max 甚至试图联系 Mark Zuckerberg，看能否获取 Facebook 的底层数据来训练模型。 此外，两人彼时都有着深度的创业创伤。TrialPay 当时正处于险些卖给 PayPal 却在最后关头“被晾在结婚礼堂门口”的痛苦阶段，Alex 心力交瘁地发誓再也不想亲自从零启动一家公司；Max 同样刚刚经历完创业退出的煎熬。因此两人的共识是搭起架子后迅速引入团队来执行。随后，Max 将 Nathan Gettings 和 Jeffrey Kaditz 介绍进来，四人正式注册成立了公司。Alex 回忆道，这一切真正的转折点和现实落地，发生在此后 2012 年的 Allen & Company 会议上，当时他们拿着写在 PHP 上的原型演示，开启了一段充满戏剧性的商业破冰之旅。 ### [27:07](https://www.youtube.com/watch?v=J3pegsM5drk&t=1627s) Allen & Company 峰会上的 1-800-Flowers 通宵 Demo 在为 Affirm 寻找最初突破口时，Max Levchin 与 Alex Rampell 查阅了 Allen & Company 峰会的参会嘉宾名单，发现了 1-800-Flowers 的创始人兼广告代言人 Jim McCann。为了在那场仅有约 200 人的精英会议上争取到一个早餐会面机会，Alex 连夜用 PHP 写出了一个完整的 Demo：他克隆了 1-800-Flowers 的电商网站，设计了一套“以身份支付”（Pay with your identity）的结账流程——场景是购买一打玫瑰花给妻子“赔罪”，在结账提示“没带信用卡”时，选择通过 Facebook 登录。 后台系统会调用 Facebook Connect，校验用户是否拥有超过 500 名好友，并借用 Facebook 当时的内部风控标记（判断该账号是否为刚注册的虚假小号，因为账号不等于真人）。 早餐会上，Jim McCann 展现出极大的共鸣。他回忆起公司早期的真实经历：常有现役军人打电话订花送给妻子，但身边没有信用卡，客服便会体谅地说“感谢您的服役，下次再付就行”。McCann 直觉上完全理解“建立在足够信任基础上的后付款”逻辑；当 Max 和 Alex 表态愿意全额承担用户赖账的坏账风险时，McCann 极度兴奋，当场拍板同意合作，并自此成为了 Max 职业生涯中长期坚定的支持者。 - **克隆 Demo 的技术细节**：Alex 独自通宵用 PHP 完整仿制了 1-800-Flowers 的选花结账闭环，以“没带卡但可用社交身份验证结账”为核心卖点。 - **风控逻辑**：利用 Facebook 好友数（门槛 500+）及 Facebook 当年的反机器人/虚假账号识别标签作为初步身份与信用凭证。 --- ### [29:25](https://www.youtube.com/watch?v=J3pegsM5drk&t=1765s) 拍脑袋的 7% 费率与“免费送花”的嘲弄 接手具体落地的业务负责人是 1-800-Flowers 的高管 Amit Shah（日后也是一名成功的创业公司创始人）。在推进商业化合作时，当时团队甚至没有正规的财务人员，Alex 便在 Word 或 Excel 中临时攒了一份定价单，拍脑袋定出了 7% 的商户费率（MDR，Merchant Discount Rate）。消费金融中包含商户扣率（MDR）与年化利率（APR）两个维度；由于 Affirm 当时先给商户垫付全款、日后再向消费者催缴且不向用户计息，实际上就是 APR 为零、商户承担 7% MDR。 这份商业模型在给早期财务人选 Rob Feifer（其姐姐曾在 Max 手下工作）看时，遭到了无情嘲笑。Rob Feifer 一针见血地评价道：“这不就是免费送花吗（Free flowers）？”因为消费者当时根本没有任何违约惩罚或必须还款的强制约束，早期的坏账率也确实相当惨淡。尽管被调侃，团队仍顶着质疑前行。此时公司的注册主体仍为“Expedite Software Inc.”，不久后才正式更名为 Affirm。 ```mermaid flowchart LR accTitle: Early Affirm Flow accDescr: Early Expedite / Affirm merchant and customer transaction dynamic. subgraph User["消费者端"] U1["购买商品（如鲜花）"] --> U2["Facebook 身份验证"] U2 --> U3["0% APR / 后续付款"] end subgraph Affirm["Expedite / Affirm"] A1["垫付 100% 货款"] A2["向商户扣除 7% MDR"] A3["承担所有坏账风险"] end subgraph Merchant["1-800-Flowers 商户端"] M1["获得 93% 现金回款"] M2["承担更高交易费率"] end U1 --> Affirm Affirm --> M1 A2 --> M2 ``` --- ### [30:57](https://www.youtube.com/watch?v=J3pegsM5drk&t=1857s) 穿越“大迷茫”：漏斗底部的内耗与商户指责 Max 指出，几乎每家日后成功的创业公司都要经历一段犹如在“荒漠中游荡 40 年”的至暗时刻（The 40 years in the desert）。在这个阶段，每天去公司上班虽有名义上的意义，但随时可能以彻底的一无所获（Great nothing）告终。1-800-Flowers 上线后，Affirm 正好深陷这一泥潭：尽管产生了一点点交易量，但对接负责人 Amit Shah 极度挑剔且不满，持续批评产品的转化率低下和 UI 体验差。 更致命的是产品逻辑的缺陷：Affirm 当时仅作为结账页面底部的一个付款选项出现（漏斗下游）。商户对此怒不可遏：“你们根本没有为我带来增量，只是在蚕食我原有的信用卡交易！而且信用卡手续费才多少，你们居然收我 7% 的 MDR？如果免费我当然愿意接，收这么贵你们简直应当感到羞耻！”此时的 Affirm 既不能给商户拉新，费率又显著高于 Visa/Mastercard，陷入了“虽然创始人团队声誉在外，但产品毫无价值”的严重停滞期。 --- ### [33:00](https://www.youtube.com/watch?v=J3pegsM5drk&t=1980s) Beautylish 破局：从“解决睡袍痛点”转向“预算扩容” 转机发生在一个名为 Beautylish 的在线美妆电商平台，其创始人 Nils 是团队的朋友。在接入产品（此时已更名为 Affirm）时，Beautylish 尝试了一种颠覆性的改动：不再只把分期付款藏在最终结账页，而是在用户挑选洗发水、香水的**漏斗上游（商品展示页与详情页）**直接标注——“可分为 3 期付款”或“支持 30 天后付款”。 这一上游透传改动带来了惊人的爆发力，页面转化率瞬间暴增 30%。团队由此迎来了真正的顿悟时刻（Aha moment）：**Affirm 的本质从来不是解决“没带信用卡”的睡袍痛点（Pajama problem），而是解决“消费者当前预算有限，但若分期偿还其购买力可大幅扩张”的消费金融需求**。 一旦掌握了这个核心公式，Affirm 迅速以此为标杆发起销售攻势，立刻引爆了一大批处于爆发期的 D2C（Direct-to-Consumer）小品牌： - **转化率与提成不在乎**：这些新兴品牌极度渴望拉动营收天花板（Top line），只要能带来爆发式销售，商户根本不在意收取的费率是 1%、5% 还是 12%。 - **Tradesy 的狂热支持**：二手奢侈品电商 Tradesy 的创始人 Tracy 成为早期超级铁杆，频繁把后台数据截图发给 Max，激动地标注“这就是 Affirm 效应，你们让我们的转化率整整飙升了 35%”。 - **DTC 床垫品牌的协同浪潮**：随后迎来的是 Casper、Purple 等互联网“盒装床垫”品牌的井喷。这些高毛利、大客单价品类由于具备强大的毛利空间，极为契合分期金融服务。 - **高 MDR 的极端案例**：Alex 进一步补充，商户愿付极高 MDR 的行业往往是营利性教育机构（如 Apollo Group 旗下的凤凰城大学 University of Phoenix），其线上课程毛利率极高，即便预期多数学生违约，也愿意向垫资方让渡高达 50% 的 MDR，将坏账与应收账款风险完全推给资金方。 ### [35:54](https://www.youtube.com/watch?v=J3pegsM5drk&t=2154s) 盒装床垫的爆发：极高毛利与转化率的突破 Alex Rampell 和 Max Levchin 探讨了 DTC（Direct-to-Consumer，直接面向消费者）床垫品牌的崛起。床垫不再是过去笨重的传统大件，而是被压缩并卷成一个小圆柱体或装进箱子里送货上门；拆封后就像泡水玩具一样瞬间膨胀复原。这种创新的交付形态背后有着惊人的商业逻辑：床垫由石油提炼而成的记忆棉制成，制造成本极低，毛利率高达 80% 甚至更高。 极高的毛利率让床垫商家拥有充裕的利润空间来承担高额的商家折扣率（MDR，Merchant Discount Rate）。更重要的是消费者心理：深夜躺在不舒服的旧床垫上穿着睡衣刷手机，用户很难狠下心一次性掏出 1,200 美元买一张新床垫；但如果将其拆解成按月分期的小额付款，转化率就会出现爆发式提升。床垫不仅成为了 Affirm 攻克的代表性场景，更是随后大规模进军整个 DTC 品牌生态的跳板。 ```mermaid flowchart TD accTitle: DTC床垫高毛利驱动分期付款闭环 accDescr: 展示DTC床垫品牌如何通过高毛利和降低心理门槛促成Affirm零利率分期的商业逻辑。 subgraph CostMargin["生产与利润结构"] A["低成本记忆棉压缩包装<br/>（盒装/圆柱体送达）"] --> B["毛利率高达 80%+"] end subgraph MerchantIncentive["商家激励机制"] B --> C["承受更高 MDR 费率"] C --> D["贴补利息提供 0% APR 分期"] end subgraph ConsumerConversion["消费者转化效果"] D --> E["降低 $1,200 决策阻力"] E --> F["睡衣场景下转化率激增 30%+"] end ``` - **超高毛利率**：盒装床垫的生产本质上是“将石油转化为泡沫塑料”，生产端高度集中，毛利率高达 80% 以上。 - **痛点匹配**：消除深夜消费者面对千元美金大额消费的心理压力，通过月付拆分实现极大幅度的转化率拉升。 --- ### [36:47](https://www.youtube.com/watch?v=J3pegsM5drk&t=2207s) 误入在线教育的教训与退场 在全面拥抱 DTC 品牌之前，Affirm 曾短暂尝试过为在线教育机构（包括类似于 General Assembly、编程训练营以及各类 University of Phoenix 式的在线培训）提供学费支付方案。当时团队认为，教育机构愿意补贴学费是与 Affirm 普惠金融的品牌使命相吻合的，然而 Affirm 最终在短短半年内“尖叫着仓皇逃离”了该赛道。 退出原因在于此类教育产品的质量普遍堪忧。学员在结课后发现自己拿到的证书或学到的技能一文不值，心生懊悔并开始拒绝还款。这种因消费者对课程质量极度不满而产生的违约率（Loss Rates）居高不下，直接证明了缺乏核心价值支撑的劣质服务无法靠消费金融模型跑通。 - **业务尝试**：为 6 周速成编程营及在线培训机构提供学费贷款解决方案。 - **失败诱因**：学员发现培训缺乏实质价值导致满意度极低，平台坏账率过高，Affirm 在短短 6 个月内果断终止该领域合作。 --- ### [38:00](https://www.youtube.com/watch?v=J3pegsM5drk&t=2280s) 七年置换周期引发的获客白热化 话题重新回到 DTC 模式与垂直整合业务。Alex 与 Max 讨论了床垫制造的集中化现象（如犹他州的生产集群，以及 Casper、Purple 等品牌的涌现）。这场创业浪潮很大程度上由当时一篇《哈佛商业评论》（HBR）的文章所引爆：文章指出消费者的床垫更换周期平均为 7 年，如果一家品牌错过了用户的当前换新节点，就必须再等 7 年。 这种“非买即失”的紧迫感，加上床垫本身极高的毛利，促使 4 到 5 家明星创业公司同台竞技，掀起了不计一切代价抢夺消费者的营销大战。为了抢先拿下用户或将换新周期缩短至 6 年，商家甘愿承担任何获客成本。提供“由 Affirm 支持的 3 年期零利率分期”不仅成为强有力的促转化工具，更成为品牌核心的营销招牌，为 Affirm 提供了让商家出资补贴息费、向用户呈现真实 0% APR 的完美土壤。 - **核心行业数据**：消费者床垫平均更换周期为 7 年，单次转化成败决定品牌近十年的销售窗口。 - **极度获客驱动**：品牌方愿意通过承担 Affirm 的 MDR 来支持长达 36 个月的免息贷款，以此作为差异化获客手段。 --- ### [40:37](https://www.youtube.com/watch?v=J3pegsM5drk&t=2437s) 怒斥“虚假 0%”：消除滞纳金与递延利息的陷阱 Max 详细分享了驱动 Affirm 商业伦理的核心动力：对传统信用卡行业“伪 0%（Fake 0%）”把戏的深恶痛绝。在传统的百货商场联名信用卡中，广告上虽然写着“0% APR”，但旁边通常带有一个不起眼的星号标记（`*`）。这些小字条款属于“递延利息（Deferred Interest）”机制：如果消费者在 12 或 24 个月的周期内少还了一分钱或逾期了一天，前期所有免除的利息都会从消费第一天起全额追溯回补，原本 1,000 美元的消费可能两年后变成 3,000 美元的巨额账单。 Affirm 抓住了广大消费者被这种欺诈性条款伤害的痛点，确立了绝对透明的产品原则：Affirm 的“0% 就是真正的零利率”。无论用户晚还款一个月还是一年，平台绝对不收取滞纳金（Late Fees），绝不引入递延利息，绝不搞暗藏的猫腻与隐形陷阱。这种“绝不让用户在负面意外中被坑”的承诺，成为了 Affirm 在市场上迅速赢得口碑并吸引所有主流床垫商家跟进的关键。 | 维度对比 | 传统零售商联名信用卡 | Affirm 分期支付 | |:--- |:--- |:--- | | **0% 宣传属性** | **伪 0%（带星号 `*` 的递延利息）** | **真 0%（零利率、透明展示）** | | **违约与逾期处理** | 只要欠一分钱或晚一天，利息溯及既往自起始日全额追加 | 绝不追溯收取惩罚性利息，无滞纳金 | | **费用透明度** | 依靠隐形陷阱、滞纳金与违约罚款盈利 | 前端明码标价，总费用锁死，无额外 Gotcha 扣费 | --- ### [42:15](https://www.youtube.com/watch?v=J3pegsM5drk&t=2535s) 从满足需求到创造需求：支付与广告的终极融合 随着 Affirm 的发展，市场往往将其简单归类为“先买后付”（BNPL，Buy Now Pay Later），但 Max 指出这一标签严重低估了公司的本质。早在 Affirm 创立之初的邮件讨论中，团队就提出过“支付与广告正在走向融合”的核心论点。这一判断虽然承袭自 Alex 早期的 TrialPay 以及 Max 在 PayPal 时期未竟的“PayPal Shops”项目（因超前时代 15-20 年而受阻），但在今天已完全成为现实。 Affirm 最初是帮助商家“满足需求”——当消费者手头预算有限、在鞋子和包包之间犹豫不决时，Affirm 通过透明条款或零息分期促使其全部买下。如今，Affirm 仅在全美就已经积累了超过 5,000 万真实交易用户，并拓展至四个国家。 Affirm 正在演变成一个帮助品牌发布新品、触达核心受众并“创造与锁定需求”的高效平台，真正完成了从单纯交易工具向广告营销阵地的跃迁。 - **用户规模**：仅在美国本土即拥有超过 5,000 万交易用户，并向全球四个国家及更多市场快速扩张。 - **商业本质演进**：从在结账环节“转化已存在的购物需求”，升级为依托海量用户消费偏好为品牌“定向创造新需求”。 --- ### [44:05](https://www.youtube.com/watch?v=J3pegsM5drk&t=2645s) “负获客成本”（Negative CAC）的商业壁垒 Alex Rampell 从风险投资人的视角指出，当前 90% 的消费类创业公司本质上都在给 Google 和 Meta 打工，因为其大部分资金都被高昂的用户获取成本（CAC）吞噬，以至于买这些公司的股票还不如直接买 Google 或 Meta 的股票。 在此背景下，Affirm 展现出了极为罕见且强大的商业模式特征——“负用户获取成本”（Negative Customer Acquisition Cost）。由于 Affirm 能为商家带来巨大的销售转化提成，商家不仅在结账页面将 Affirm 放在显眼位置主动推广给千万用户，还会直接向 Affirm 支付高额服务佣金。Affirm 不但无需为获取终端消费者支付高昂的广告费，反而在每一次“获取新用户”的过程中都在从商家端赚钱。这种自带网络效应与负 CAC 的底层结构，是常规白标（White-label）支付工具所无法企及的壁垒。 ### [44:46](https://www.youtube.com/watch?v=J3pegsM5drk&t=2686s) 负获客成本（Negative CAC）与品牌赋权的天然护城河 Alex Rampell 剖析了 B2B2C（面向企业的企业再面向消费者）商业模式的复杂性，并指出了 Affirm 与众不同的核心优势。Alex 以自己早年创立 TrialPay 的经历作为鲜明对比：当年用户在 Zynga 玩游戏时，如果注册了 Geico 车险，TrialPay 会赠送游戏金币；但消费者根本不知道 TrialPay 是何方神圣。当 TrialPay 尝试直接给这些用户发送电子邮件时，甚至遭到了邮件服务商的封禁，因为虽然在法律协议上拥有联络权，用户心理上却毫无品牌认知，而且像 Zynga 和 Netflix 这样的合作商家坚决不希望第三方插手拥有属于他们的客户。 但在 Affirm 的模式下，商家的态度完全相反——他们**巴不得**由 Affirm 来拥有和维系这层消费者关系。以床垫品牌 Casper 为例（Alex 幽默地拿卡通形象“小鬼卡斯伯”和自己童年养的一只白猫打趣），当消费者发生逾期或还款违约时，Casper 绝不想亲自给顾客寄催收信，大喊“赶紧从我的床垫上滚下来，你还没付钱！”品牌需要第三方金融机构充当缓冲层。对于 Affirm 而言，这意味着他们能够在实现“负获客成本（Negative CAC）”的同时，与 5000 万消费者建立起直接且稳固的金融契约关系。Alex 强调，很多风投支持的公司虽然能做到 Casper 那样的业务规模，但其背后堆积着巨额的获客成本（CAC），随着所有经济价值逐渐被 Google 和 Meta 等广告巨头吸干，商业模式往往会随时间退化；而拥有负 CAC 和真实客户掌控权的企业，几乎是风投界极难得一见的奇迹。 ```mermaid flowchart TD accTitle: Affirm's Negative CAC and Brand Separation Model accDescr: Diagram contrasting merchant customer experience burden with Affirm's direct customer financing relationship. subgraph Merchant["商户端体验 (如 Casper)"] M1["专注核心业务 / 品牌建设"] M2["摆脱售后催收与 12-39 期账单催促负担"] end subgraph AffirmRel["Affirm 平台枢纽"] A1["负获客成本 (Negative CAC)"] A2["建立 5000 万真实金融账户连接"] A3["12 至 39 次持续互动的射门机会 (Upsell)"] end subgraph Consumer["消费者生命周期"] C1["单笔大额消费分期购买"] C2["按期偿还账单与信用互动"] C3["复购与拓展至全平台新金融产品"] end M1 -->|将账务关系移交| AffirmRel AffirmRel -->|直接管理还款与信审| Consumer Consumer -->|持续还款与高粘性| AffirmRel ``` - **商户端的情感负担**：商家本身已经背负着产品售后、差评处理和技术支持的沉重负担，面对需要向用户发送 12 次甚至长达 3 年半共计 39 次还款通知的琐碎流程，商户非常乐意将催收与管理责任完全交由 Affirm 承担。 - **负获客成本的战略价值**：商家在结账页面主动为 Affirm 提供流量并贴补费用，Affirm 不仅不花钱买量，反而能在掌控客户关系后，零边际成本地向其推销更多新金融产品。 --- ### [47:00](https://www.youtube.com/watch?v=J3pegsM5drk&t=2820s) 长期信贷的壁垒：机器学习与 39 次“射门机会” Max Levchin 深入解释了 Affirm 与其他短期“先买后付”（BNPL）竞品的底层差异。市场上大多数同行只敢做平均周期为 6 周的“四期免息（Pay-in-4）”，而 Affirm 则是全行业极少数能够规模化承揽长达 3 年半（42 个月）长期无抵押贷款的机构。要在这么长的时间周期内控制违约率和逾期率，没有任何捷径可走——传统的 FICO 信用分在这里完全失灵，去社交网络翻查 Facebook 好友更是无稽之谈，唯一的护城河是极其前沿、精密的机器学习风控模型。 这种对硬核底层风控的坚持带来了两大独特优势：其一是极高的竞争壁垒，其他对手根本做不到且难以模仿；其二则是长期的用户生命周期价值（LTV）。Max 将其形象地比作“12 次到 39 次的射门机会（shots on goal）”。在用户长达数年的逐月还款过程中，每一次账单推送、每一次还款交互，都是向其推荐新产品或服务的黄金触点。尽管管理长期贷款的高风险与复杂性是整个支付行业中最艰难的一环，但这恰恰是驱动 Affirm 商业飞轮持续转动的根本燃料。 --- ### [48:18](https://www.youtube.com/watch?v=J3pegsM5drk&t=2898s) 拆解“PayPal 黑帮”神话：野心、招募机制与见证凡人的一面 谈及支付行业的历史，主持人提到了 Jimmy Soni 所写的传记《The Founders: The Story of PayPal and the Entrepreneurs Who Shaped Silicon Valley》，并询问 Max：为什么当年的 PayPal 会拥有如此惊人的“人才命中率”，孕育出硅谷历史上绝无仅有的“PayPal 黑帮”（PayPal Mafia），其后不仅诞生了 YouTube、Yelp、LinkedIn，还走出了彼得·蒂尔（Peter Thiel）与埃隆·马斯克（Elon Musk）等巨擘？ Max 对此给出了两个维度的剖析。第一个是众所周知的制度化原因：**从创立之初，PayPal 的面试流程就在刻意筛选创业者**。当时面试团队和晚期候选人时，Max 常问的一个标准问题是：“离开 PayPal 之后你打算干什么？”如果对方回答“这是我的最后一站，之后我要自己去创立一家公司”，这反而是最受青睐的标准答案。PayPal 内部宛如一所终极的“创业学研究生院”，汇聚了一群极其聪明、极具野心且志在必得的狠角色。 第二个维度则是 Max 极少对外分享的个人深刻体会：**他们曾毫无保留地见证过彼此作为“凡人”的一面**。早期核心团队挤在汗流浃背的狭小会议室里，站在白板前激烈争吵、彼此破口大骂对方的想法极其愚蠢。Max 熟悉他们卸下公众伪装后的真实模样： - **埃隆·马斯克（Elon Musk）**：外人看到的是立志带领人类移民火星的孤胆英雄；但 Max 曾无数次在公司茶水间里，看到他精疲力竭、大汗淋漓、满面愁容又牢骚满腹的普通人状态。 - **彼得·蒂尔（Peter Thiel）**：公众眼中深不可测的哲学思考者；但 Max 曾接到彼得在融资旅途中陷入巨大自我怀疑时打来的绝望电话，听他在电话另一头说：“我觉得我们的钱可能真的要烧光了。” - **大卫·萨克斯（David Sacks）**：如今在政界与播客界呼风唤雨的策士，当年同样身处极度高压的泥潭中。 Max 感慨道，当你在年轻、脆弱且充满怀疑的岁月里，朝夕目睹过身边这些“普通人”的焦虑与挣扎，日后看到他们去挑战那些看似不可能的宏伟目标时，你心中绝不会产生对“神灵”的畏惧，而是会油然而生一种坚定的自信：“既然这个我熟得不能再熟的普通哥们儿都能去干这么大的事，那我也能去闯一闯。”正是这种在汗水与挫折中锤炼出的野心与知根知底的底气，让这群人在 PayPal 散伙之后，以惊人的密度在硅谷各处开花结果。 --- ### [52:44](https://www.youtube.com/watch?v=J3pegsM5drk&t=3164s) 代理商贸的迷局：看淡“AI 替你挑衣服”，押注“代理支付”重塑信用卡 在节目的最后，当被问及还有什么未被探讨的核心观点时，Max 给出了一个反主流直觉的判断：**对于“AI 代理代客购物（Agentic Shopping）”持怀疑态度，但对“代理支付（Agentic Payments）”抱有极高的期望。** ```mermaid flowchart LR accTitle: Agentic Shopping vs Agentic Payments accDescr: Max Levchin's view on why Agentic Shopping will fail in discretionary goods while Agentic Payments will transform card infrastructure. subgraph Shopping["代理购物 (Agentic Shopping) - 低预期"] S1["AI 代替挑选个性化商品"] -->|丧失消费乐趣与决策主权| S2["用户本能抗拒 (缺乏信任)"] end subgraph Payments["代理支付 (Agentic Payments) - 强看好"] P1["传统塑料信用卡 (60年固化界面)"] -->|AI 代理介入资金路由与信贷优化| P2["终结静态卡片垄断"] end ``` Max 认为，科技界许多关于“机器人将全权帮我们挑选周五晚会穿搭”的描摹完全走偏了方向。人类进行消费——尤其是主观性、非标性的消费时，参与挑选本身就是乐趣的核心。Max 自嘲自己是个自行车狂热爱好者，经常花大量时间坐在电脑前反复比对两款细微差异的自行车零件，他绝不会把这种挑拣的快乐让渡给机器人。 然而，在支付与交易履约的流程中，情况截然相反。长久以来，“塑料信用卡”被公认为是商业历史上设计最成功的用户界面（UI），即便后来植入了可读写芯片，其底层逻辑也从未被真正撼动。但如今，AI 代理在处理信息路由、资金流匹配和信贷选择上的智能水平，已经远远超越了一张冰冷的塑料卡片。Max 总结道，当下业内关于代理商贸（Agentic Commerce）的讨论偏离了至少 20 度的航向：与其费尽心机去说服用户信任机器人挑选商品的眼光，不如将 AI 代理的强大算力倾注于重塑数十年未变的信用卡支付流程。 ### [53:44](https://www.youtube.com/watch?v=J3pegsM5drk&t=3224s) 代理化商务的雏形：时间与金钱的权衡与比价逻辑 对话从支付界面的未来演变展开。Max Levchin 认为，过去的支付交互往往聚焦于交易的最终环节，而如今则是从消费者拿出钱包、甚至更早的环节就开始切入，支付的用户界面正迎来一轮极为深刻的创新。 Alex Rampell 提出了一个延展问题：消费者购物可以分为“调研买什么”和“落地去买”两个阶段。在调研阶段，AI 只是众多工具之一，比如想买一个自行车配件，用户既会向懂行的朋友（如 Max）请教，也会去问 AI。然而一旦确定了具体的 SKU（统一商品编号/UPC），用户根本不在乎最终是从哪家电商把它买回来的。此时世界的消费者大致可划分为两类：更在乎时间的人，以及更在乎金钱的人。Alex 认为，“代理化商务”（Agentic Commerce）的本质并不是让 AI 替你决定买哪个汽水或哪个配件，而是让用户抛出一个明确的商品型号后，发出指令：“去帮我用最低的成本把它买回来。” Alex 特别提到了比价监控网站 CamelCamelCamel（亚马逊历史价格跟踪工具），称其为绝佳佐证：这是全美排名前 100 的网站之一，但在年收入超过 10 万美元的人群中几乎无人知晓，因为该群体普遍不愿花费大量时间去紧盯折扣。对于手头时间远多于金钱的人来说，跨网站比价早已是基本操作；而代理化商务与代理化支付的结合，正是把这一特定人群原本依靠人力完成的繁琐工作，自动化并推广给更多人——类似于手握 9 张信用卡的人，根本记不住哪张卡在非餐饮、低于 500 美元的消费上返现比例最高，如果 AI 能代为优化这种微观决策，将带来巨大的实际价值。 ```mermaid flowchart TD accTitle: 代理化商务决策路径与用户心智模型 accDescr: 展现从挑选商品到最终交付中时间、金钱与信任的权衡 subgraph Research["1. 商品调研阶段"] A["明确购买意图"] --> B["多方咨询与对比<br>(朋友建议 / AI问答)"] B --> C["锁定具体 SKU / UPC"] end subgraph Preference["2. 用户心智画像"] C --> D{"核心权衡倾向"} D -- "时间敏感型<br>(Time > Money)" --> E["优先可靠度与速度<br>不愿费心微调卡片返现"] D -- "价格敏感型<br>(Money > Time)" --> F["极致追求全网最低价<br>(如 CamelCamelCamel 逻辑)"] end subgraph AgenticExecution["3. 代理化执行挑战"] E & F --> G["AI 代理寻找货源与最优卡片"] G --> H{"履约不确定性风险"} H -- "信誉存疑 / 物流慢" --> I["消费者退缩并选择大牌平台"] H -- "履约透明可靠" --> J["全自动购买与交付"] end ``` ### [55:44](https://www.youtube.com/watch?v=J3pegsM5drk&t=3344s) 履约信任鸿沟与长尾电商的“90年代网站”难题 Max Levchin 对此指出，尽管从方向上看“代客寻找最低价并完成支付”逻辑合理，但它的实际采纳曲线（Adoption Curve）可能会比大家预期的慢得多。虽然科技圈常说“先是缓慢演进，随后突然爆发”，但目前普通消费者的真实境遇并非简单地二元划分为“只要时间”或“只要金钱”，大多数人处于某种中间状态——购买决策的精细程度与商品品类、客单价息息相关。比如买牛奶这类日用必需品，人们直接随手买下；但如果是客单价昂贵的专业自行车飞轮（Cassette，一套可能售价 600 到 1000 美元），消费者的考量就会完全不同。 Max 强调，消费者不愿单纯交给 AI 比价下单的核心阻碍在于**履约风险的概率评估**（Handicapping Probability）：即便远方某个不知名的小零售商给出的价格最便宜，消费者也无法确定该商品能否准时送达、是否为全新正品、还是齿轮已被磨损的二手翻新件。面对三个同款商品的卖家，人们通常会牺牲价格去选择声誉最好的那家。Max 调侃称，一个有经验的用户只要肉眼扫一眼网站界面，心里就会嘀咕：“这网页看起来像是在 20 世纪 90 年代做的，我真的敢在这里掏 1000 美元买配件吗？”Alex 随即笑着补充，自己想到的正是像“Bike Closet”这类页面复古却卖极低价飞轮的网站。Max 认为，当前的 AI 或许已经具备抓取低价的能力，但普通用户还没有建立起足够的信任，无法放手让 AI 代替自己去权衡信誉、时效与潜在退货风险之间的动态变化。 ### [57:48](https://www.youtube.com/watch?v=J3pegsM5drk&t=3468s) 已经成型的代理化消费：Instacart 与未被察觉的未来 在展望未来时，Max 话锋一转，提出了一个令人意外的积极视角：其实**完全代理化的消费（100% Agentic Commerce）在现实中已经发生**，而且每天都在被数以百万计的人使用，那就是以 Instacart 为代表的代客买菜平台。 用户在使用 Instacart 时，对采购代理（Shopper）完全授权——用户只要写下“帮我带一瓶有机牛奶”，如果用户原本指定的 Organic Valley 缺货，代购员换成了另一个品质相当的品牌并直接结账送达，用户会觉得“理所当然，完全合情合理”。Max 调侃道，即便代购员大脑里的即时算力与判断力未必比得上当下的 Gemini 或 ChatGPT，但用户早已经完全接受并信任这种模式：只要把需求丢给对方，让他们在 Trader Joe's 或 Whole Foods 穿梭挑选，最终送到家门口的商品有 99.9% 都能符合预期甚至超出预期。 这种心智模式证明，人类已经完全适应了将“选品、支付与物流履约”全流程彻底外包。阻碍更广泛品类实现代理化商务的关键，不是算法不会比价，而是如何妥善解决后续的现实摩擦：比如买错了自行车配件如何退货、包裹在途失踪后由谁负责等履约短板。一旦这些现实阻力被逐步克服，代理化购物与代理化支付的全面落地只是时间问题。主持人 Erik Torenberg 随即总结，这无疑是极佳的收尾切入点，并期待在未来的第二期播客中进一步深聊金融与 AI 的交汇前景。 ## Notable quotes > [00:00](https://www.youtube.com/watch?v=J3pegsM5drk&t=0s) 信用卡支付界面被公认为“有史以来创造出的最优秀的单一用户界面（UI）”，而整个支付领域也毫无争议地构成了全球最庞大的市场。在支付赛道中，根本不存在规模小于 1000 亿美元的所谓“利基市场”。 > [02:13](https://www.youtube.com/watch?v=J3pegsM5drk&t=133s) 身处这个行业 30 余年，最不可思议的事情其实是底层规则的顽固不化。自 Dee Hock 在 20 世纪建立 Visa 体系以来，卡组织的底层硬性规则已经沿用了整整 60 年，至今仍未被撼动。 — Max Levchin > [05:54](https://www.youtube.com/watch?v=J3pegsM5drk&t=354s) 交易金额越大，实际产生的费率空间反而越小。 > [22:08](https://www.youtube.com/watch?v=J3pegsM5drk&t=1328s) 你这辈子工作最拼命、但内心最快乐的日子，就是当年在 PayPal 做反欺诈系统的时候。那时虽然累到近乎虚脱，但你眼里有光。 — Max Levchin 的妻子 > [29:25](https://www.youtube.com/watch?v=J3pegsM5drk&t=1765s) 这不就是免费送花吗（Free flowers）？ — Rob Feifer > [30:57](https://www.youtube.com/watch?v=J3pegsM5drk&t=1857s) 你们根本没有为我带来增量，只是在蚕食我原有的信用卡交易！而且信用卡手续费才多少，你们居然收我 7% 的 MDR？如果免费我当然愿意接，收这么贵你们简直应当感到羞耻！ — Amit Shah > [33:00](https://www.youtube.com/watch?v=J3pegsM5drk&t=1980s) Affirm 的本质从来不是解决“没带信用卡”的睡袍痛点（Pajama problem），而是解决“消费者当前预算有限，但若分期偿还其购买力可大幅扩张”的消费金融需求。 > [44:05](https://www.youtube.com/watch?v=J3pegsM5drk&t=2645s) 当前 90% 的消费类创业公司本质上都在给 Google 和 Meta 打工，因为其大部分资金都被高昂的用户获取成本（CAC）吞噬，以至于买这些公司的股票还不如直接买 Google 或 Meta 的股票。 — Alex Rampell > [48:18](https://www.youtube.com/watch?v=J3pegsM5drk&t=2898s) 既然这个我熟得不能再熟的普通哥们儿都能去干这么大的事，那我也能去闯一闯。 — Max Levchin > [55:44](https://www.youtube.com/watch?v=J3pegsM5drk&t=3344s) 这网页看起来像是在 20 世纪 90 年代做的，我真的敢在这里掏 1000 美元买配件吗？ — Max Levchin ## Mentioned Max Levchin, Alex Rampell, Erik Torenberg, Dee Hock, David Chaum, Jim McCann, Amit Shah, Rob Feifer, Nils, Nathan Gettings, Jeffrey Kaditz, Mark Zuckerberg, Peter Thiel, Elon Musk, David Sacks, Jimmy Soni, Affirm, PayPal, TrialPay, Slide, Google, Apple Pay, Google Pay, Visa, Mastercard, Europay, DigiCash, 1-800-Flowers, Beautylish, Tradesy, Casper, Purple, General Electric, Starbucks, Bill Me Later, Facebook, FarmVille, Geico, Netflix, Amazon, Whole Foods, Mobil Speedpass, Instacart, CamelCamelCamel, Bike Closet, University of Phoenix, General Assembly, The Founders: The Story of PayPal and the Entrepreneurs Who Shaped Silicon Valley ## Video description Erik Torenberg is joined by a16z General Partner Alex Rampell and Affirm Co-Founder and CEO Max Levchin for a conversation on 25 years of fintech, from the early days of digital payments to the origins of Affirm and the next generation of agentic commerce. Max and Alex revisit what surprised them most about how payments evolved, why the card interface has been so difficult to displace, and why even the smallest corners of payments can become enormous markets. They also trace the early idea maze behind Affirm, from "pay with your identity" and the pajama problem to the realization that installment financing could dramatically increase merchant conversion. The conversation also gets into real versus "fake" 0% financing, what people misunderstand about Affirm today, why negative customer acquisition cost can be such a powerful business model advantage, and why Max is more bullish on agentic payments than on agents choosing what people buy. Timestamps: 00:00 - Intro 01:01 - Meet Max & Alex: 25 Years of Thinking About Fintech 01:33 - What's Surprised Them: Apple Pay, Google Pay & Changing Consumer Behavior 03:36 - Max on the 60-Year-Old Rules Still Running Payments 10:00 - What Idea Should Exist But Doesn't Yet? 12:36 - Did Crypto Feel Like the Realization of a Dream, or Orthogonal? 17:15 - The Origin of Affirm: Solving the Pajama Problem 20:01 - Pay With Your Identity: Social Credit & the General Store Model 27:00 - The Allen & Company 1-800-Flowers Demo 30:57 - The 40 Years in the Desert & Finding Product Market Fit 33:00 - Beautylish, Mattresses & the 30% Conversion Breakthrough 42:15 - What People Miss About Affirm Today: From Satisfying Demand to Creating It 48:58 - The PayPal Mafia: Why So Many Founders Came Out of It 52:44 - Agentic Payments vs Agentic Shopping: The Real Innovation Ahead Resources: Follow Max Levchin on X: https://x.com/mlevchin Follow Alex Rampell on X: https://x.com/arampell Stay Updated: If you enjoyed this episode, be sure to like, subscribe, and share with your friends! Find a16z on X: https://twitter.com/a16z Find a16z on LinkedIn: https://www.linkedin.com/company/a16z Listen to the a16z Show on Spotify: https://open.spotify.com/show/5bC65RDvs3oxnLyqqvkUYX Listen to the a16z Show on Apple Podcasts: https://podcasts.apple.com/us/podcast/a16z-podcast/id842818711 Follow our host: https://x.com/eriktorenberg Please note that the content here is for informational purposes only; should NOT be taken as legal, business, tax, or investment advice or be used to evaluate any investment or security; and is not directed at any investors or potential investors in any a16z fund. a16z and its affiliates may maintain investments in the companies discussed. For more details please see http://a16z.com/disclosures. [Source](https://www.youtube.com/watch?v=J3pegsM5drk)

[0:00] 信用卡支付界面被公认为“有史以来创造出的最优秀的单一用户界面（UI）”，而整个支付领域也毫无争议地构成了全球最庞大的市场。在支付赛道中，根本不存在规模小于 1000 亿美元的所谓“利基市场”。 [2:13] Max Levchin: 身处这个行业 30 余年，最不可思议的事情其实是底层规则的顽固不化。自 Dee Hock 在 20 世纪建立 Visa 体系以来，卡组织的底层硬性规则已经沿用了整整 60 年，至今仍未被撼动。 [5:54] 交易金额越大，实际产生的费率空间反而越小。 [22:08] Max Levchin 的妻子: 你这辈子工作最拼命、但内心最快乐的日子，就是当年在 PayPal 做反欺诈系统的时候。那时虽然累到近乎虚脱，但你眼里有光。 [29:25] Rob Feifer: 这不就是免费送花吗（Free flowers）？ [30:57] Amit Shah: 你们根本没有为我带来增量，只是在蚕食我原有的信用卡交易！而且信用卡手续费才多少，你们居然收我 7% 的 MDR？如果免费我当然愿意接，收这么贵你们简直应当感到羞耻！ [33:00] Affirm 的本质从来不是解决“没带信用卡”的睡袍痛点（Pajama problem），而是解决“消费者当前预算有限，但若分期偿还其购买力可大幅扩张”的消费金融需求。 [44:05] Alex Rampell: 当前 90% 的消费类创业公司本质上都在给 Google 和 Meta 打工，因为其大部分资金都被高昂的用户获取成本（CAC）吞噬，以至于买这些公司的股票还不如直接买 Google 或 Meta 的股票。 [48:18] Max Levchin: 既然这个我熟得不能再熟的普通哥们儿都能去干这么大的事，那我也能去闯一闯。 [55:44] Max Levchin: 这网页看起来像是在 20 世纪 90 年代做的，我真的敢在这里掏 1000 美元买配件吗？

[7:15] Max Levchin: There are no niches in payments that are smaller than a hundred billion. [12:36] Max Levchin: The big innovation of PayPal was: what if we don't care about anonymity at all? In fact, people don't. They just need to pay for their coffees or their online purchases.

[0:00] The card payment [music] interface is [0:02] the singular best user interface ever [0:04] created. It is the world's largest [0:05] market [music] by any stretch of [0:06] imagination. And there are no niches and [0:08] payments that are smaller than hund00 [0:09] billion. [0:10] >> Once you go really big, the numbers get [0:12] small, which is strange. [music] There's [0:14] a lot of volume, but the large volume [0:17] revenue opportunities and payments tend [0:19] to be the smaller dollar amounts. [0:20] >> There's always an opportunity to use [0:22] another form of payment delivery device [0:25] to satisfy a basic need. convenience [0:27] just trumps as the total amount you're [0:30] trying to send goes down. The best user [0:31] interface ever created is the credit [0:33] card. This may [music] actually be [0:35] finally up for renegotiation because AI [0:37] is already there. It's just that you [0:38] haven't yet trusted your agent to do as [0:41] good a job as you would. There is [0:42] something else that you're surprised has [0:43] not happened yet. I went to some [0:45] cryptography related conference and [0:47] presented a new idea in digital payments [0:49] and was literally booed off stage [0:51] because it was certainly not anonymous [0:53] and the the big innovation of PayPal was [0:54] what if we don't care about anonymity at [0:56] all and in [music] that sense um [1:01] so you guys go way back having co-ounded [1:03] a firm and and met well before then you [1:05] guys are pioneers in in fintech you've [1:08] been thinking about the space trying to [1:10] make sense of the present you know think [1:11] about where the future is going for 20 [1:14] plus years, 25 years, maybe even more. [1:17] And I'm curious, you know, given so much [1:20] time has passed since you first got into [1:21] the space, what has most surprised you [1:24] about what has happened, what hasn't [1:26] happened, what did you expect to happen [1:28] in the early 2000s of how how this space [1:31] would would play out? Maybe Alex, why [1:33] don't you start? I think the rise of [1:35] Apple Pay and Google Pay and to the [1:36] extent the the extent to which they've [1:38] really penetrated because it's very hard [1:40] to change consumer behavior in general [1:42] and it was a bizarre set of accidents if [1:45] you will where there is this merchant [1:46] liability shift because magst stripes [1:49] were so easy to replicate that uh you [1:52] probably noticed a while ago you your [1:54] new cards had a little chip on them and [1:56] you had to dip the chip and then the [1:58] machine would be like you know don't [1:59] remove don't remove right that's because [2:01] the chip was being rewritten it was much [2:03] more secure than a magstripe and [2:05] basically Visa and Mastercard. Um, and [2:07] then the Euro Pay that's what the E is [2:09] for. Uh, but what is EU for? EMV. [2:13] >> Europe card. [2:15] >> Euro card. [2:16] >> I don't know. The Europeans somehow got [2:17] their say because they always do. They [2:19] they insert their not electronic. That's [2:21] >> Yeah, it's it's euro something [2:23] Mastercard Visa. So, it was it was [2:24] called the EMV switch. So, all of these [2:26] machines, if you did not want to be [2:28] liable, um, like I could go to Best Buy, [2:31] buy a TV, and then get home, use my TV, [2:33] and say, "Nope, never bought the TV." [2:35] And if I was using my magstripe and Best [2:37] Buy didn't have the the machine switch [2:39] over, then Best Buy would be like, "Oh, [2:40] shoot. We lost all the money on that TV, [2:42] and Alex gets the free TV." I'm [2:45] exaggerating slightly for a fact. So, [2:47] everybody had to get a new machine. And [2:49] then it turned out that those new [2:50] machines had a little contactless, you [2:53] know, thing where you didn't have to dip [2:55] the chip or swipe the card. You could [2:56] just tap. Nobody was using taps when [2:59] those machines came out and now it's [3:01] ubiquitous. So that's that's something [3:02] where I mean generally speaking it's [3:05] hard to change consumer behavior but I [3:07] feel like between COVID and the fact [3:09] that all these merchants independently [3:11] had to switch over their machines and [3:13] the fact that I mean I wouldn't have [3:14] guessed in I mean you obviously did [3:16] because the uh confinity was was Palm [3:19] Pilot sending money. So you you're ahead [3:21] of the whole PDA wave but I wouldn't [3:23] have guessed that those three things [3:24] would have come behavior. It's like [3:25] change of of consumer behavior, [3:28] ubiquitous new merchant payment terminal [3:30] plus ubiquitous uh mobile telefan and [3:34] that's that's certainly one thing that [3:35] has changed the world a lot. [3:37] >> How about you Max? You you were dreaming [3:39] about you know what what the future of [3:40] money would look like a long time ago. [3:42] Would it have surprised you about what [3:44] has happened or what hasn't happened? [3:45] It's very hard to uh step back from the [3:49] soup if you're cooking in it uh for the [3:52] last plus or minus 30 odd years. [3:56] Uh, I think the to sort of further [4:01] illustrate Alex's point, one of the [4:03] things that's really amazing and subtle [4:05] about Apple Pay/Google Pay and its [4:09] interaction with the Visa Mastercard [4:11] networks is Visa and Mastercard have a [4:15] hard 2 and a half second limit on [4:18] interaction between the network, the [4:20] issuing bank and the merchant and the [4:22] acquiring bank. And so the whole thing [4:24] has to happen in 10 seconds where the [4:25] transaction gets retrieded or maybe just [4:27] canceled out and really leaves very [4:30] little room for any kind of clever [4:32] innovation. And online you can play [4:34] games. If you're doing e-commerce you're [4:35] like well yes we're going to submit your [4:37] card to Visa but first we're going to [4:39] run some anti- fraud checks and we're [4:40] going to do some other things to like [4:41] reduce liability. [4:43] But offline like once your card is [4:44] presented two and a half seconds that's [4:46] all you got. And Google and Apple Pay [4:50] have singularly timeshifted the whole [4:52] thing by creating secure enclaves inside [4:54] their chips and saying I already know [4:56] your card. I can do all kinds of things [4:58] before I actually have to talk to Visa [4:59] Mastercard. And maybe the shocking thing [5:01] is that Visa Mastercard have not yet [5:03] introduced some new standards saying [5:04] actually it doesn't have to be two and a [5:05] half seconds at all. It could be 15 [5:07] seconds while we know go and get a bunch [5:09] of issuers to bid on a better credit [5:11] quality terms for you or any other type [5:14] of innovation. [5:15] the sort of hardwritten rules of dehawk [5:18] era are more or less intact and that's [5:21] not 30 years that's like more you know [5:23] 60 years and so I think that's probably [5:26] a critical or more criticizing take on [5:29] what happened um [5:32] but the shocking thing about payments is [5:34] that it is the world's largest market by [5:36] any stretch of imagination and [5:39] anything and everything you could have [5:40] possibly thought of being sort of [5:42] deconstructed and being the nichiest [5:44] little thing to poke around and innovate [5:46] in always turns out to be a hundred [5:48] billion dollar. There are no niches and [5:50] payments that are smaller than hundred [5:51] billion. And so that's probably the most [5:53] surprising thing I've seen. [5:54] >> Yeah. [5:54] >> Although it's interesting like once you [5:56] go really big, the numbers get small, [5:59] which is strange, right? Like if I send [6:01] a $1 trillion wire, right? The United [6:03] States just announced we we are $40 [6:05] trillion in debt, right? Let's just say [6:08] that like, you know, Elon's Mars colony [6:11] is worth $40 trillion. and you know [6:13] Roana decides to tax him $40 trillion [6:15] and he needs to send a wire that wire is [6:19] not going to be very pro like there [6:20] there's a lot of volume but the large [6:23] volume revenue opportunities and [6:25] payments tend to be the smaller dollar [6:27] amounts um which is surprising right [6:30] because it turns out I mean like all of [6:32] the QSR the quickserve restaurant the [6:35] payment opportunity there that's massive [6:37] that's why Starbucks invented Starbucks [6:39] pay because they wanted to stop having [6:40] to pay it's like these prepaid they only [6:42] have to pay one time. You char you put [6:44] $50 on your Starbucks account. Well, [6:46] they only have to pay Visa, Mastercard, [6:47] and everybody else one time as opposed [6:50] to every successive time. So, a lot like [6:52] what's surprising about payment uh area [6:55] under the curve, if you will, is that [6:57] like the bigger the dollars, like the [6:59] smaller the rake, which I guess makes [7:00] sense because you're not going to get 2% [7:02] of that $40 trillion payment, but [7:04] everybody goes after B2B payments. [7:05] Everybody has a clever idea for B2B [7:07] payments and like that's the only one [7:09] that is probably the exception to the [7:10] rule. I was trying to find so it's lost [7:13] in time the original email between the [7:15] two of us which I think we have slightly [7:17] different origin stories of how we [7:18] actually met [7:19] >> but I was trying to find the original [7:20] email between the two of us where we [7:22] start mentioning what becomes a firm [7:24] eventually and it's dated [7:27] April 2011 [7:29] and uh in it [7:32] I think I ask you like tell me more [7:34] about this bill me later thing because I [7:36] was working on something else while you [7:38] were watching PayPal buying build me and [7:40] you send me these really long elaborate [7:42] sort of here's exactly what it is and [7:44] all these ideas around it and we then [7:46] veer off into this ping pong of emails [7:49] around bill me later for business and [7:51] conclude that it's actually a dumb [7:53] business where there's there isn't an [7:54] opportunity [7:55] >> well and that's actually it's funny you [7:56] mentioned I bought the domain pay me [7:58] sooner [7:59] >> as a result so it's like ah bill me [8:01] later was a good idea pay me sooner [8:02] would be an even better idea actually [8:04] the idea there it still is it's not it's [8:06] not terrible because it's not payments [8:07] it's lending so think of it this way [8:09] Imagine that. Like big companies beat up [8:12] small companies all the time to get [8:14] better payment terms. Better payment [8:16] terms for them, right? It's like so if [8:18] I'm GE and you're Eric Torrenberg and [8:20] you're a little guy, I'm big GE. I'm [8:22] like, I'm going to pay you net 90. So [8:24] you invoice me and I pay you 90 days [8:26] later. Now you have to pay your [8:28] employees. So you go to a bank, you go [8:30] to somebody else. You're like, hey, I'm [8:32] Eric. Uh, I run this business. Most of [8:34] my money comes from GE. They're an [8:36] awesome company. They're going to pay me [8:37] in 90 days. bank doesn't care, right? [8:40] They're just like, "Yeah, we're going to [8:41] charge you 15%." Meanwhile, GE can issue [8:44] bonds at like, you know, whatever sofur [8:47] plus 10 basis points or something. Like [8:49] GE is paying 5% for credit. You're [8:52] paying 15% for credit, but it kind of [8:54] doesn't make sense because your credit [8:56] is G's credit, if that makes sense. Now, [8:58] there's something called factoring where [9:00] you can sell the invoice. You can [9:01] actually sell the receivable. That's [9:03] somewhat usurious. um that that's very [9:06] very actually it's not usurious because [9:07] it doesn't like credit laws don't apply [9:09] because you're selling a receivable but [9:11] I always thought that pay me sooner was [9:12] an interesting idea because the whole [9:15] economy runs on the fact that like my [9:17] business is waiting on payments from you [9:20] like the reason why I have to borrow [9:21] money is because you don't pay me sooner [9:24] uh or other way around sorry I don't pay [9:25] you sooner and yeah so but yes we didn't [9:28] we didn't pursue that [9:29] >> there there's some healthy businesses [9:30] these days in both accounts payable [9:32] financing and accounts receivable [9:33] financing so it's [9:34] This is it's not a bad idea, but the [9:37] revenue opportunity despite being an [9:39] unregulated lending space is seemingly [9:41] lower than the one in consumer because [9:44] the convenience factor is something that [9:47] every participant understands and there [9:49] are plenty of players that'll offer you [9:50] a slightly cheaper chart which is also [9:52] the root origin of why a $40 trillion [9:54] wire to Roana is going to be uh [9:57] >> not very profitable business whoever [9:58] processes it. [9:59] >> Yes. What what's what's another idea [10:02] that you're surprised uh does not exist [10:04] yet that you were thinking at some [10:05] point, hey, the world obviously should [10:07] work this way? There's, you know, an [10:09] obvious opportunity here and for [10:10] whatever, you know, conflation of [10:12] reasons just hasn't happened yet or or [10:13] to use uh you know, Mark's quote, there [10:15] are no bad ideas, only early ideas. Is [10:18] is there something else that you're [10:19] surprised has not happened yet? [10:23] You know, we're forever in search of the [10:25] great way to pay biometrically, and it [10:28] manifests itself more in movies where [10:31] people's thumbs and eyeballs get removed [10:33] for authentication purposes. [10:36] And yet, like, we're still paying with [10:38] chips. By the way, the card payment [10:40] interface is the singular best user [10:42] interface ever created. And it it's not [10:45] something that people haven't tried to [10:47] do better. remember even before PayPal, [10:50] there was this wand that I think [10:52] Mastercard and one of the gas um [10:57] uh refueling station networks produced [10:59] this thing where you could like wave [11:00] this wand next to a gas station and [11:02] payment is secured and you can fuel your [11:04] car. And when I saw it, I'm like, "Oh my [11:05] god, this is obviously going to take [11:07] over credit cards." And it didn't. And [11:09] there actually therein lies a really [11:11] interesting lesson in payments [11:13] innovation in general. There's some [11:16] unobvious a priority but clear [11:20] opposeriori version of critical mass and [11:22] if you don't reach it you're going to [11:24] fail like there's not really a okay [11:26] outcomes in payments. You either get [11:28] there and everyone needs to have your [11:30] widget or your network or your whatever [11:33] or it's just going to go into the uh the [11:36] time. And this is a fine example of a [11:38] thing that actually made a lot of sense. [11:39] Like of course you want to have a little [11:41] wand on your car keys to just sort of [11:43] wave and get refueled, but it's just a [11:45] little bit faster, not a lot faster than [11:46] the credit card you have in your pocket. [11:48] And that just works. And so I I keep on [11:50] waiting for some completely different [11:52] way of authenticating the payer to the [11:55] device and like other than the mobile [11:57] phone, we haven't done better. [11:59] >> Amazon actually just discontinued the [12:01] whole thumb payment at Whole Foods. [12:03] >> They actually It's not a thumb payment. [12:04] It's like it's the palm. The palm. Yeah. [12:06] >> I loved it. I I would literally go to [12:09] Whole Foods next to whatever hotel I'd [12:11] be staying in. [12:12] >> I was sad that they got rid of it. I [12:13] know. I used to use it, too. [12:15] >> It's actually not even faster. It's just [12:16] fun. [12:17] >> I know. It's probably slower. [12:19] >> It is slower. It's like tell me my [12:20] fortune and these grapes. [12:22] >> The fortune is you owe me money. Yes. [12:23] That is that is the only fortune that [12:25] you ever get. Well, that that was I mean [12:27] I think I called it thumb by by uh by [12:30] accident just given the the almost uh [12:32] almost name of a firm. [12:34] >> That's right. [12:36] when when sort of the crypto industry [12:38] was was becoming popular or when you [12:40] know some of the major projects uh were [12:41] were starting in in the space did that [12:44] were there parts of that that seemed [12:45] exciting to you or realization of a you [12:47] know long-held uh sort of dream or did [12:49] it feel hey orthogonal or just not [12:52] really [12:54] >> so I I'm frequently late to trends like [12:57] I'm not the world's earliest adopture of [13:00] majority of things I encounter [13:01] >> I feel like PayPal is pretty early right [13:04] Well, [13:05] sort of. Um, to give you full context [13:08] for PayPal, I really wanted to build [13:12] something very specific and quite [13:13] different from PayPal uh that operated [13:15] on very low power chips. And to do that, [13:18] I needed to work out how to make [13:21] cryptographic primitives work on a very, [13:23] very low power chip. And we did. And [13:28] then we promptly realized that it really [13:29] couldn't do very much other than encrypt [13:31] some very very small amounts of data and [13:34] small amounts of data that need to be [13:35] encrypted and decrypted quickly that [13:36] kind of leads you to payments which is [13:38] how PayPal came about. But right before [13:40] all of that began I went to the [13:44] windown/bank bankruptcy party for [13:46] DigiCash which was the original [13:49] granddaddy of all digital payments. And [13:52] uh it was sort of a somber affair. I was [13:54] on Stanford University grounds and there [13:57] was like you know pouring out of the 40s [13:59] and a bunch of very sadl looking cyber [14:01] punks telling each other how the age of [14:03] digital payments is not upon us and uh [14:06] so in that sense we were very late when [14:08] sort of I was listening to these people [14:10] talk whose idea was crushed it was blind [14:13] signatures and David Ch was somewhere he [14:15] was actually not present but he was [14:17] roaming the uh streets of Palo Alto [14:19] despondent because his brilliant idea [14:21] and a truly brilliant one was you know [14:23] basically being proven to be not having [14:25] a a product market fit. Uh and then we [14:28] promptly soon thereafter started PayPal [14:30] and uh I went to some cryptography [14:33] related conference and presented you [14:35] know a new idea in digital payments and [14:38] was literally booed off stage because it [14:41] was not nearly as secure. It was [14:42] certainly not anonymous and the the big [14:44] innovation of PayPal was what if we [14:45] don't care about anonymity at all. In [14:47] fact, people don't. They just need to [14:48] pay for their coffees or their, you [14:50] know, online purchases. And um in in [14:53] that sense, um I feel like we were again [14:56] very late to that game. The game had [14:58] played out like all the enthusiasts had [15:00] left the building. Uh and so [15:03] cryptocurrency, [15:04] not to be confused with cryptography. [15:07] uh when so I I read the original Bitcoin [15:10] paper and thought, "Wow, that's a really [15:11] clever way of solving the Byzantine [15:13] general problem. I sort of as all the uh [15:16] math cryptography nerds was impressed [15:18] with the approach, but I not think for a [15:21] second it would become a currency or a [15:23] payment method and etc. And I'm not [15:26] convinced it's a payment method now and [15:28] it's been quite some time, but as [15:30] currency and commodity and a store of [15:32] value goes, it's certainly proven to be [15:34] extraordinarily successful." So I think [15:35] the [15:38] as I watched uh cryptocurrencies [15:42] expand and and become a thing and more [15:43] and more of a thing, the use cases are [15:45] coming out now. We're, you know, stable [15:47] coins are upon us and that's certainly a [15:49] collection of very clear uses, but none [15:54] have, in my opinion, come close to [15:57] breaching the [15:59] I'm going to use this to buy a cup of [16:00] coffee, which is kind of that's the [16:02] canonical and most important, I would [16:04] argue, use case for old forms of [16:07] payments. So, as currencies go, as store [16:10] value goes, fascinating. [16:12] You know, when was the last time you [16:13] spent a Bitcoin or a Satoshi on anything [16:17] actually important? You know, putting [16:18] aside people who make it a point of [16:20] spending it just to sort of prove the [16:22] point wrong [16:24] >> and and it's the most important um that [16:26] buying coffee is the most important [16:27] metric because of just the the volume of [16:29] it because the frequency, the practical, [16:32] >> there's always an opportunity to use [16:34] another form of payment delivery device [16:37] to satisfy a basic need. So, if you're [16:39] sort of contemplating a $40 trillion [16:41] transfer, you'll spend a lot of time [16:42] figuring out the secure way, the the [16:44] fast way, the cheap way, etc. If you're [16:46] passing through your nearby bagel shop [16:48] and you need a cup of coffee, if your [16:51] cryptocurrency wallet passphrase is too [16:53] long, you're going to look for change in [16:56] your pocket. And if you don't have it, [16:57] you'll pull out a debit card or credit [16:59] card. So user interface as the payment [17:02] amount diminishes it takes over [17:07] cost takes over everything. Basically [17:10] convenience just trumps everything else [17:12] as the total amount you're trying to [17:14] send goes down. [17:15] >> Yeah. [17:17] I I want you to take us both back to me [17:19] memory lane. It's the beginning of a [17:20] firm and how you guys navigated the idea [17:23] maze of of what would eventually become [17:26] a firm. [17:28] Um, [17:30] you go here. [17:30] >> Well, hopefully our stories match. [17:32] >> Yeah, we should put us in three [17:33] questions for like no separate route. [17:35] This this is like the prisoner's [17:36] dilemma, right? It's just like ask if [17:38] our stories match, we have to we get to [17:40] leave. If they don't match, then we have [17:41] to stay here all day. Um, so my [17:44] recollection was um I ran this company [17:46] Trial Pay, which did um alternative [17:49] payments for digital goods. So, you [17:51] don't want to pay for throwing a sheep [17:54] or doing something on some silly social [17:56] game, something that was not hard, [17:57] valuable, or fun. [17:58] >> Who would do such? [17:59] >> Who would who would do such a thing? [18:00] But, you know, you're buying coins in [18:01] Farmville or you're doing something. [18:03] You're buying poker chips, virtual poker [18:04] chips. Of course, you don't want to pay [18:06] get it for free if you sign up for [18:07] Geico. There's a lot of economic value [18:09] to Geico. You don't really care if you [18:11] use Progressive or Geico. Oo, I can get [18:13] Farmville coins if I switch to Geico or [18:15] get a credit card or sign up for [18:16] Netflix. So, that's what we did. And uh [18:18] I I was uh somewhat persuasive and [18:20] annoying. Probably more annoying than [18:21] persuasive. So signed up every other [18:23] social gaming company. Um except for [18:26] this one called Slide. And uh Max ran [18:30] proudly or unproudly ran Slide. And uh I [18:33] went to the Allen & Company conference [18:34] which I almost didn't go to because my [18:36] wife was very very pregnant. So this was [18:38] in March of 2009. It's like the first [18:41] week of March. My son was born March [18:44] 28th, 2009, but her due date was March [18:46] 16th. So like I'm not going to go to [18:48] this conference. And then uh the the [18:50] Allen guys were like, "This is a very [18:52] good conference. You should really go." [18:53] I was like, "My wife's very pregnant." [18:54] He's like, "You should go to the [18:55] conference. If she goes into labor, [18:57] we'll fly you back." I was like, "All [18:58] right, that sounds good. I'm going." So [19:00] I went [snorts] to the conference. I [19:01] think we met there briefly. Uh [19:04] successfully did not persuade you. U but [19:06] then I think I followed up with a note [19:08] in Russian. [19:08] >> Yes. Um, and this wouldn't have worked [19:11] today because now obviously I would have [19:13] used Gemini or you know you know chat [19:15] GBT to write this probably somewhat [19:17] riddled with grammatical flaws Russian [19:19] note actually that that probably would [19:20] have given it away of not being chat [19:22] GPT. It wouldn't have passed it would [19:23] have passed panggram but I took Russian [19:25] in in high school and college worked [19:27] there lived there for a while um and I [19:30] think you were like why do you speak [19:31] Russian? I think that was the response [19:32] but then we just became friendly. Uh I [19:34] think I you you sold to Google so I [19:37] could not ever convince slide to become [19:39] a client of of trial pay sadly. Um but [19:43] uh I I seem to recall you had like a [19:46] tweet like I'm looking for something to [19:48] do. Any idea? And that was like my [19:49] reason to reach out and I think we we [19:51] had coffee at the forbidden building at [19:53] Google. Remember how like there was one [19:54] Google building that no guests were [19:56] allowed at but you just didn't care. Um [19:58] statute of limitations were [laughter] [19:59] so they can't. [20:00] >> You're revealing all my secrets. [20:01] >> Yes. Exactly. Because I know how much [20:02] you love Vic. Um, so, so, [laughter] so [20:05] we we had coffee there and then that's [20:07] where this Yeah, we were talking about [20:08] Bimley later, but it was like, you know, [20:10] my recollection was it's really hard to [20:12] pay for something on a mobile phone. [20:13] Mobile phones are becoming more [20:14] ubiquitous. We should have a way of [20:17] solving the pajama problem of you're [20:18] upstairs in your pajamas, you want to [20:20] buy something, your credit card is [20:22] downstairs, how do you pay? Meanwhile, [20:24] not just are mobile phones ubiquitous, [20:26] but social networks are ubiquitous. And [20:28] maybe there's a new underwriting [20:29] mechanism of kind of going back to the [20:31] old general store from like the 1800s of [20:33] it's like you show up in the general [20:34] store, you don't have cash, like both [20:37] literally and figuratively. You don't [20:38] have it on you, but you don't even have [20:40] it, you know, back home. But you run the [20:42] general store and you're like, "Okay, [20:43] Max, I got you. Don't worry about it." [20:45] But today, you're just like a cookie and [20:48] an IP address, so I don't really know [20:49] who you are. Or if you go to Walmart [20:51] offline, it's like, you know, you get [20:53] greeted by the very nice greeter. I [20:54] don't think they have them anymore, but [20:55] like you get greeted by the nice greet. [20:56] they have no idea who you are. You can't [20:58] do the general store thing. But if you [21:00] have 500 friends on Facebook, this was [21:02] wrong, by the way, but if you have 500 [21:03] friends on Facebook and they all have, [21:04] you have a thousand pictures that you've [21:06] uploaded, you're probably a low credit [21:08] risk. And if you're not looking for [21:09] credit, this is the key thing. It's like [21:11] a lot of credit offers are sent out [21:14] proactively. So, you have an 800 FICO. I [21:17] know that you have $14,000 that's [21:19] revolving for some reason on your [21:21] Capital One card. I'm going to send you [21:23] a customized mailer saying, "Why don't [21:24] you go refinance with me, Bank of [21:26] America?" But if you go on Google and [21:28] you say like, "I'm out of money. Need [21:30] money, need money, credit, credit, [21:31] credit," like you're probably a bad [21:33] credit risk. So, kind of going back to [21:34] the general store concept, people that [21:36] aren't looking for credit that might not [21:39] have their wallet with them, kind of [21:41] make it like the 1800s and pay with your [21:44] identity. That that was kind of how I [21:45] remembered it. Yeah, I think that that [21:47] that matches. Uh the [21:51] to add a little bit more color to it. So [21:55] the year at Google was definitely [21:58] a challenge in a sense that I was still [22:00] coming to terms that I ran for like over [22:03] 5 years a company in social media that I [22:05] had no business running. Like I I worked [22:07] very hard. [22:08] had a financially very positive outcome [22:10] for a lot of people, but ultimately it [22:12] never sort of scratched the sort of [22:14] entrepreneurial itch that I thought I [22:16] was I was scratching. And so one of the [22:18] things that happened during that year, [22:21] I spent a lot of time soulsearching like [22:22] what am I supposed to do next with my [22:24] life? I was going to start a company, [22:25] but you know, maybe maybe I should be [22:26] more thoughtful about what company I'm [22:28] going to start next. And my wife, who's [22:30] pretty much always right, said, "You [22:32] know, [22:34] the hardest you've ever worked and the [22:36] happiest you were was during the [22:38] anti-fraud days at PayPal, like when you [22:41] nearly died, or we nearly died. She was [22:43] there as well." Um, [22:46] those were the days when you looked [22:48] exhausted, but you actually were [22:50] obviously very happy. And you know, I [22:53] you know, she was telling me, "I know [22:54] you swore off doing financial services. [22:56] I know you're never going to touch [22:56] payments again, but just this once. Give [22:59] give it a chance. And that that actually [23:00] had some influence as to my [23:02] responsiveness, too, because you're [23:03] working on a payments company. And part [23:04] of my reluctance to deal with trial pay [23:07] was like, I can't handle another near [23:10] payments company. Like I I went into [23:12] social media for a reason. I don't want [23:13] anything to do with this stuff. And so, [23:14] as I was slowly getting reeled back into [23:18] working on payments, [23:20] we started talking about this. The sort [23:21] of a social BML was one of the sort of [23:24] short hands we used. discussing what it [23:27] would be. Um, the general store analogy [23:31] was interesting because at the same [23:33] time, I can't remember which one of us [23:34] knew this cuz you also speak Japanese [23:36] and this this could have been the thing [23:37] you told me, but I remember reading [23:39] about social credit in different [23:41] economies. There's a notion of just pay [23:45] me later or pay me next time in Israeli [23:47] grocery stores, which is still common [23:48] today. If you're checking out an Israel, [23:50] apparently in a small enough town, you [23:52] can just show your bag to the cashier. [23:54] they'll remember and then they'll get [23:55] you next time if you forgot your money. [23:57] This is even more formalized in Japan [23:59] where until very recently you would give [24:01] your business card to the storekeeper [24:04] even if they didn't know you. If you had [24:05] a business card, they would write what [24:07] you bought on the back of it and that [24:09] was your total. And the notion of like [24:10] you just put it on my tab is a thing [24:12] that's you know obviously fairly [24:13] American concept and so on. And so we [24:16] were converging on this. Uh I was [24:17] talking to a startup at the time. [24:21] ascouch just learning about the industry [24:23] that was trying to build a social credit [24:25] score kind of modeling it on the Chinese [24:27] social score which is used for a lot [24:30] more than than credit obviously. So we [24:32] were swirling around this idea of [24:36] payments and solving the I don't have my [24:38] credentials with me but I still need to [24:40] get through this. Like another sort of [24:42] metaphorical version of this was I'm [24:44] watching TV. did this really cool ad [24:47] back in the day when people had [24:48] advertisements in between things in TV. [24:50] Uh, and I really want the thing that was [24:53] promised to me at the price that it was [24:54] promised to me, but I could do it on my [24:57] phone, but I'm definitely not going to [24:59] get off my bed. And so, like what if [25:00] there's some area under the curve that [25:02] you can capture by bringing these um [25:06] transactions closer in time and [25:08] improving conversion obviously. And so, [25:09] that that's the kind of the the swirly [25:11] origin story. Um, [25:15] a lot of it I think the two of us had a [25:17] slightly different version of what would [25:18] be fun. So I think you as a payments guy [25:22] you were mostly thinking like let's go [25:23] do some payments stuff. As a person [25:26] recovering from doing lots of machine [25:28] learning in a service of social media [25:30] which was you know good and well but [25:33] wasn't really what I was meant to do and [25:35] lusting for the days of doing machine [25:37] learning to fight fraud. My mo was I [25:41] just want to build a really cool credit [25:42] score and then someone else will take [25:45] care of payments like lending payments. [25:47] Don't really know anything about it. And [25:48] I've done payments. Don't need to do [25:50] lending. But building a really cool [25:51] score that would be amazing because I [25:53] remember just the reams of data we had [25:56] access to at PayPal. We could maybe get [25:58] some data. Maybe we could buy some data. [25:59] We could mine Facebook for data. I'm [26:01] trying to get in touch with Mark [26:02] Zuckerberg to see if he might give me [26:04] some of his data. And so a lot of that [26:06] was sort of in my head while you were [26:07] like I just want to close some [26:09] transactions. I think that that maybe if [26:11] there's any difference between the [26:12] motivations you were much more motivated [26:14] by let's go find a merchant. [26:16] >> Well but the similar motivation was like [26:18] you know trial pay I think we were in [26:20] the throws of selling to PayPal but then [26:21] they they left me at the altter not your [26:23] PayPal the many many generations later [26:25] PayPal. Um and I was like I never want [26:28] to start a company again. This is so [26:30] painful. I and then you you had been in [26:32] that in the throws of that as well. So, [26:34] it was also like let's get this thing [26:36] going. Um, and we'll hire other Well, [26:40] but then you introduced me to Nathan uh [26:42] and Jeff um and we kind of formally [26:45] incorporated, but I I also like the [26:47] other thing that I distinctly remember [26:49] and you'll you remember this in a very [26:51] negative light because it's PHP, but uh [26:53] we had a meeting this I I kind of think [26:55] of the origin as being the Allen and [26:57] Company conference in 2012. [26:59] >> Oh, yeah. I I don't I don't find it a [27:02] negative thing. [27:02] >> No, no, I know. But it was funny. So, so [27:04] we have this meeting like so, uh, we're [27:06] going through the the list with, uh, [27:07] son, remember your your old chief of [27:09] staff, right? So, we're going through [27:10] the list and it's like, oh, we both know [27:12] Jim McCann, who's this very friendly guy [27:14] who used to be on all the commercials [27:15] for 1800 Flowers. Remember 1800 Flowers, [27:17] you can order flowers. Order it right [27:19] now for me or whatever the commercial [27:21] was. So, uh, we ping him to get a [27:22] breakfast meeting. It was like 200 [27:24] person conference. Anybody can be with [27:26] anybody, but it's like, hey, why don't [27:27] we have breakfast? And I I pulled like [27:29] an allnighter making this demo um in [27:32] PHP. Yeah, I basically like cloned. It's [27:34] just funny like a I could do this in 5 [27:35] seconds now, but I I I cloned 1800 [27:37] Flowers site um and then made a whole [27:40] like pay with your identity. So you do a [27:42] checkout like add the dozen roses to, [27:45] you know, make up for the bad stuff that [27:47] I did. Sending it to my wife, click pay. [27:49] Oh shoot, I don't have my my my payment [27:51] card with me. I can pay with Facebook. [27:54] What does that do? It does Facebook [27:56] connect. It makes sure that I have over [27:58] 500 friends. And actually Facebook had I [27:59] mean to your point, they had a lot of [28:00] interesting data. They had their own [28:02] internal flags around did they think [28:04] this account was fraudulent or not. And [28:05] fraudulent for back then was more of [28:08] like it was just created not as a real [28:09] account. It wasn't a real person because [28:12] an account is not a person, right? [28:14] Hopefully they're the same thing, but [28:15] not necessarily because one person can [28:18] have multiple accounts. Um, but if [28:20] there's one to one. So, so that was uh [28:22] that was our fun demo and then I [28:24] remember Jimmy was like, "This is great. [28:25] Let's do it." [28:26] >> Yeah. So, I [28:28] two other colorful points from that. So, [28:30] it was a good breakfast. And then in the [28:33] middle of it, he goes, "Oh, yeah. We [28:34] used to do exactly this thing back in [28:36] the day when the service members would [28:39] call us and say, "Send me my wife some [28:40] flowers." And uh, "No, I don't have my [28:43] credit card with me." And like, oh, that [28:45] that's fine. Like, you know, thank you [28:46] for your service. We'll we'll just get [28:48] you on the next one. And he like [28:50] intuitively grocked this idea that like [28:52] you can absolutely postpay a thing if [28:54] you have enough trust. And we were [28:56] basically there raising our hands saying [28:57] we'll take the responsibility for, you [28:59] know, if they don't pay you, it's okay. [29:00] We're going to eat it. And uh so he was [29:03] like instantly smitten. He he had been a [29:07] great proponent of the product and like [29:10] a supportive, you know, person [29:12] slashpresence in my professional life [29:15] ever since that breakfast. And he had [29:16] absolutely no reason to trust us other [29:18] than like he has some degree of sense [29:19] who we were. And it was like you guys [29:21] are great. You seem like you really know [29:23] what you're doing. Like he was so [29:24] excited. We have no idea. [29:25] >> Well, then we got handed to uh this what [29:27] was his name? He actually now runs a [29:29] successful startup. The uh the guy that [29:31] was Amit. Yes. Amit Sha. [29:33] >> Yes. [29:34] >> Yeah. So, Amit took over. We implemented [29:37] he was very very excited. But and I [29:40] remember I I I came up in Microsoft I [29:41] think I did this in Microsoft Word or [29:43] Excel like here's our pricing sheet. We [29:44] need a pricing sheet and I was like why [29:46] don't we charge 7%. [laughter] [29:48] um like so in in consumer finance for [29:52] these things there's a concept of an MDR [29:54] and an APR. So merchant discount rate [29:56] because you're discounting what you pay [29:57] the merchant because you're advancing in [29:59] the cash right now. You're not [30:00] collecting it until later. So call it a [30:03] 7% merchant discount rate but I think [30:04] there was like some table. I just [30:06] literally made this up. Um and then and [30:08] then there was an APR. Um but the APR [30:11] here there was no APR. It was just MDR. [30:13] Um but then nobody like I I showed this [30:15] to somebody. I think it was Rob Feifer [30:17] actually uh before we hired [30:18] >> you had like a real finance employee. [30:20] >> Uh yeah. So but but I I showed him this. [30:22] He was like free flowers, right? Because [30:23] he's like you just didn't have to pay us [30:25] back. Um so the the returns were not [30:28] fantastic on that. [30:28] >> So cynical. [30:29] >> Um but uh yeah, I think he literally [30:32] said free flowers. It's like this is [30:33] this is a great deal free flowers. [30:34] >> This is just a generally a very smart [30:35] but very cynical guy. So we were not [30:38] going to be perturbed by some cynicism. [30:40] >> I love route. [30:40] >> Uh his sister used to work for me. So I [30:43] remember uh interviewing him in the [30:45] context of a firm although I think at [30:47] the time it was still Expedite. [30:49] >> Yeah. It briefly bumped around as [30:50] expedite and then we finally renamed it. [30:52] >> What was it? It was it was incorporated [30:54] as Expedite Software Inc. [30:56] >> Yeah, I think that's right. [30:57] >> And so how did the idea evolve as you [31:00] you know went from there? When did it [31:02] start to really get get product market [31:03] fit or how did the space play out in the [31:05] way that you know [31:06] >> every startup [31:07] has the 40 years in a desert like it's [31:10] just a a given like if it's worth [31:12] talking about expost I mean some some [31:14] people just like quit too early or you [31:17] know something happens to to destroy it [31:18] but I've never run into a company where [31:21] people start a company and they just [31:23] like hit product market fit and 24 hours [31:26] later they're just selling and you know [31:28] everything's going swimmingly. So you [31:29] always have this period where like kind [31:31] of still makes sense to show up to work, [31:33] but you don't really know if it's going [31:34] to end in like a great nothing. And we [31:38] were definitely meandering through the [31:39] great nothing period for a while. After [31:42] 100 flowers, we had a sort of non-stop [31:44] interactions with said Ahmed, who was [31:46] never happy with the conversion rate, [31:49] with the user interface, and he was [31:52] mostly right. like he he was kind of a [31:53] grumpy demanding guy but he knew what he [31:56] was talking about from the product and [31:58] financials perspective and he was [31:59] definitely not going to pay 7%. Uh and [32:02] so uh we were sort of limping through [32:05] that. We're generating a little bit of [32:06] volume, but among other things, the [32:11] notion of, hey, you can pay us later [32:15] showed up after you selected the [32:17] flowers. And so he would say, "Look, you [32:18] guys are cannibalizing my credit card [32:20] volume. If I charge if I were charged [32:23] nothing for it, of course, I'd send you [32:26] more transactions, but you're charging [32:27] me more than credit cards, so it's it's [32:29] a dumb idea. like I don't want to pay [32:30] anything for it at all. You know, you [32:31] guys ought to be ashamed of yourself [32:32] basically. And we were sort of trying to [32:36] convince other merchants, but we had one [32:37] merchant that was essentially saying, [32:40] "Yeah, I mean, these guys are nice guys, [32:41] but it's kind of worthless." And so it [32:43] it was it was going nowhere fast. And [32:45] then we had another friend, somebody [32:48] named Nils, who ran a company called [32:50] Beautylish, which is kind of still [32:51] around, but they were selling online [32:56] cosmetics and beauty products. And the [32:59] only difference that Beautylish [33:03] did as they implemented expedite, I [33:06] think by then it may have already I [33:07] think it was named the firm by then was [33:10] that they told their shoppers upfunnel [33:15] basically as they were selecting their [33:17] shampoos and and perfumes that you could [33:20] pay [33:22] in three installments or 30 days later [33:25] or or whatever it is we were trying to [33:26] sort of do. then and that had an instant [33:30] 30% increase in conversion and that was [33:32] the oh we know what this is this isn't [33:36] an alternative this isn't actually [33:38] solving the pajama problem it is solving [33:40] my budget is this but if I could pay [33:43] over multiple periods of time it would [33:45] expand a lot and so the second we sort [33:48] of understood the very beginnings of [33:49] that notion we turn it right around into [33:51] a sales campaign and very quickly found [33:53] a bunch of smallish brands primarily [33:57] direct to consumers who could not care [33:59] whether we charged them 1%, 5% or 12%. [34:02] Because they were trying to grow their [34:03] top line and telling their shoppers [34:05] that, hey, you could pay for this over [34:08] time instead of having to cough up the [34:09] cash right now. It was transformative. [34:11] And so I would literally get these love [34:13] letters from merchant CEOs saying, and [34:14] I'm still friends with a lot of them, [34:16] like early adopters. Most of these [34:17] people have now either sold their [34:18] businesses or the business has failed or [34:19] they've gone on to do something else. [34:21] But I still run into people who were [34:23] like there's this woman named Tracy from [34:25] Tradzy. Tradzy.com was a [34:27] >> Oh, yeah. I just saw her. Yeah. She [34:30] She's wonderful. And she was an early [34:32] advocate for us [34:33] >> because her business literally [34:35] >> I forgot about her. [34:36] >> She would email me screenshots of her [34:38] dashboards and say, "This is the affirm [34:40] effect. Here's a 35% pop you guys caused [34:43] for me. Like, please do more." [34:44] >> Well, there there was a solution to the [34:45] pajama problem. And then it was also all [34:47] the mattress companies. [34:48] >> Oh, yes. [34:49] >> That that that was [34:50] >> that was another that was the big one. I [34:52] just remember it's like, you know, wait, [34:53] there's a company called Purple and [34:55] there's a company called Casper. It's [34:56] just like they just kept coming out of [34:58] the woodwork and that was [34:59] transformative. And I remember actually [35:00] we were talking about this in the [35:02] context of like how can you charge a [35:04] high MDR and get away with it? That [35:06] sounds bad. Not get away with it, but [35:08] like actually show sufficient value to [35:09] the merchant. The highest MDRs are [35:12] actually for-profit education [35:14] institutions. Mhm. [35:16] >> So, some of them are like 50 like I mean [35:18] I I don't like these guys in particular, [35:20] but like University of Phoenix. [35:22] >> Um, so there's the private equity firm [35:24] Apollo, but there's also like this [35:25] holding company of all these for-profit [35:27] institutions called Apollo Group or [35:28] Apollo something. And University of [35:30] Phoenix, I think, is the biggest one. [35:32] And they just assume that most people [35:34] aren't going to pay. And the the gross [35:37] margins are so high because it's an [35:39] online course. They're like, "Okay, you [35:41] could take 50%." Right? you're stuck [35:43] with the receivables on the back end, [35:45] which goes back to the fraud fighting [35:46] that Max was talking about before [35:48] because the gross margins of the [35:50] mattress, like these mattresses, I don't [35:52] know if you ever bought one. It's pretty [35:53] cool because they didn't actually ship [35:54] you a mattress that looks like this. It [35:56] rolls out. So, you get this like [35:59] cylinder. It's pretty small. [36:00] >> Some of them are boxes. [36:01] >> Some of them are boxes, but some of them [36:03] it's like you just open it and then it [36:04] just like it's like one of those old [36:06] like toys that you'd put in like your [36:08] bath and then it would expand or [36:09] something. It's that was like the [36:11] mattress. So they had very very high [36:13] gross margins. So therefore a lot more [36:16] flexible on the MDR and more importantly [36:20] nobody wants like you're in your pajamas [36:21] on your uncomfortable mattress. You [36:24] don't want to spend $1,200 on a mattress [36:26] but if it's like 30 like it once you [36:28] subdivided it just has a massive [36:30] increase of conversion rate. [36:31] >> Yeah. There's a couple of other sort of [36:32] technical things that are probably worth [36:34] tapping into in in that story but that [36:36] that that was the next big leap. So, the [36:38] mattresses was the precursor to the [36:42] great expansion into DDC brands, which [36:44] I'm sure we'll touch on. But before we [36:47] get there, a firm briefly flirted with [36:51] being a payment provider to the online [36:54] educational facilities, [36:56] >> which we ran out of kicking and [36:58] screaming because the reason they're [37:00] willing to pay these enormous MDRs is [37:02] because the quality of education is [37:03] largely terrible. Yeah. So, the reason [37:05] people don't pay most of the time isn't [37:07] because they're like, "Oh, I thought [37:09] about it. Now that I have a master's [37:10] degree in basket weaving, I just don't [37:11] want to pay for it." See, it turns out [37:13] my degree is worth nothing. Like, why [37:15] did I sign up for this ridiculous thing? [37:16] And so, we were, if you go back to sort [37:19] of, you know, things like um what is it [37:22] called? General Assembly. Is that their [37:24] whole sort of class of these like we [37:26] will teach you how to code? Another PEI [37:28] thing like learn how to code. Just ask [37:30] Chad GPT to code for you. But uh back in [37:32] the day when it was still important to [37:34] learn how to code, but you didn't really [37:35] want to go to college for it. You just [37:36] want to take a six week course. There [37:38] were plenty of the sort of University of [37:39] Phoenix clones that [37:40] >> Academy. [37:41] >> Yeah. There's all the all these guys. [37:42] And uh we saw that trend and thought, [37:45] "Oh my god, like these people are [37:46] willing to subsidize education, which is [37:49] cool and on mission for us." And uh yet [37:53] the loss rates to consumer [37:56] dissatisfaction were so high we lasted [37:58] like a half a year in that space. But [38:00] back to mattresses. So the thing that's [38:02] really powerful about direct to consumer [38:05] or you can think of them as vertically [38:06] integrated businesses. So there's some [38:08] factory in Brooklyn presumably or who [38:10] knows where. [38:10] >> China. [38:11] >> China. Uh I thought the mattress guys [38:14] manufactured locally because [38:16] >> I don't know but they were all in Utah. [38:18] Remember that? [38:20] >> Yeah. I feel like it was like there was [38:21] like a Utah mattress town or something. [38:23] But I think they just all originated [38:25] there. Casper is in New York or was it [38:27] purple I think was in and then there [38:28] were like other purples that were [38:30] >> I think one of the weirder things the [38:32] manufacturing was actually all [38:34] consolidated so they ultimately sort of [38:36] paid very little for these mattresses [38:37] that were kind of all the same. I mean [38:39] don't want to reveal any industry [38:40] secrets but memory foam is memory foam. [38:42] >> Uh anyway and so they their primary [38:46] value or their their their value [38:49] equation was really interesting. So, [38:51] first of all, they bought this memory [38:52] foam in a box and sort of compressed and [38:54] shipped it to you with a cool effect. [38:56] The margin there were gargantuan because [38:58] you are in fact just sort of taking [39:00] petroleum and turning it into foam. But [39:03] the other thing is the replacement [39:05] cycle. And this actually goes to a I [39:07] think a Harvard Business Review article [39:08] that came out right before Casper was [39:10] founded. And all these guys, all these [39:11] really entrepreneurial people read the [39:13] same article which essentially said [39:15] this, people replace mattresses once [39:17] every seven years. If you're the company [39:20] that sells that mattress, [39:22] it is the most important thing in the [39:24] world. And if you miss it, the next [39:25] opportunity is seven years from now. But [39:27] the margins on mattresses even before [39:29] the sort of the foam and the everything [39:30] else are like 80% or some some [39:33] eyepopping number. And so suddenly four [39:35] or five companies were born from that [39:37] article basically saying like, "Oh my [39:38] god, so if we just figure out how to [39:40] market a mattress that you will either [39:42] shorten your replacement cycle, so you [39:44] you buy the next mattress 6 years [39:45] instead of seven years from the last one [39:47] or you just decide this mattress is the [39:49] best mattress or the coolest one or the [39:51] one that pops out of a box." that is [39:53] like oh you know there's so much money [39:54] in it and one it resulted in a [39:57] saturation of mattress industry [39:58] obviously but more to the point it [40:02] created this enormous premium on I will [40:05] just compel you to buy a mattress by any [40:06] means necessary so telling you I will [40:09] give you a three-year loan at no [40:11] interest at all powered by a firm of [40:13] course is like a small price to pay and [40:15] a cool marketing campaign and and and [40:17] and so that was there were multiple [40:20] events like this in a firm's history [40:22] where we saw an opportunity for the [40:26] merchant to significantly contribute to [40:30] the MDR which allowed us to eliminate [40:33] the APR. In other words, consumer got a [40:35] true 0% loan. The backdrop to this [40:37] that's actually really really important [40:38] to understand and one of my sort of you [40:40] know things that I rage against have and [40:42] and will continue to until it's made [40:44] illegal or eliminated by by other means [40:47] is the fake 0% law. So if you ever have [40:49] gone to a, you know, fill in your [40:52] favorite department store, there is a [40:54] sign somewhere these days somewhat less [40:56] prominent that says get a branded credit [40:59] card from fill in the blank 0% APR. [41:02] There's an asterisk next to the zero [41:04] which says something along the lines of [41:06] assuming you make a purchase for the [41:08] next 12 months or 24 months, if you're a [41:12] penny short or a day late paying your [41:15] principal, the interest acrru [41:17] retroactively to the beginning of time. [41:19] Basically, the day you got that card, [41:20] you swiped it for $1,000. You wake up [41:22] and you owe $3,000 two years later. So, [41:25] this is called the deferred interest [41:26] credit card. And a huge opportunity we [41:29] tapped into [41:31] directly from that rage was this idea, [41:33] well, our 0% is going to be a real zero. [41:35] When you go to buy the inflatable [41:37] mattress or the mattress from a box, [41:39] you're going to get a 0% loan. And even [41:40] if you're a month late or a year late, [41:43] we're not going to change the price [41:44] because we are so sick and tired of [41:46] these ridiculous people who are lying to [41:48] you about zero with an asterisk. It will [41:49] never be an asterisk on a firm zero. [41:51] That that is the origin story of why we [41:53] don't charge lead fees. We don't do [41:54] deferred interest. all the sort of [41:55] gimmicks and gotchas that the industry [41:57] loves to throw at people. We we've run [41:59] away from all of that primarily as a [42:01] point of like you know what when you [42:02] decide to do this for three years with [42:04] us you will never get screwed. You will [42:05] never be surprised to the negative and [42:07] so that that was another sort of big [42:09] bounce up point and then then every [42:11] mattress company was like oh wow like [42:13] these guys are doing it we should do the [42:15] same thing. Okay so fast forwarding a [42:16] bit a lot of people hear a firm today [42:19] and you know one quick thing they think [42:20] is you know buy now pay later. What does [42:22] this not fully encapsulate or what does [42:24] this miss about the company today? It's [42:26] in its entirety. One of the things in [42:29] the email thread that I just reread in [42:31] the origin origin of a firm was your [42:34] claim that advertising and payments are [42:36] converging which was the thesis behind [42:38] trial pay thesis behind a bunch of [42:39] different things actually a failed [42:41] project within PayPal called PayPal [42:42] shops where I was trying to personally I [42:44] built this thing where was trying to [42:46] remarket what you just bought. That [42:49] affirmation should give us a sense for [42:50] what we might buy next and you know all [42:52] that. And all these things by the way [42:53] have turned out to be true. Like there [42:55] was just you know 15 20 30 years before [42:57] their time. But a firm more and more [43:00] today is means for merchants to not just [43:04] satisfy demand. So somebody comes in and [43:06] says look I'm going to buy this thing [43:07] but my budget is this big. I won the [43:09] shoes and I love a bag but I'm and a [43:11] firm is there say actually we can help [43:14] you finance this on transparent terms. a [43:16] lot of times, a third of the time, uh, [43:18] no interest at all. You should get the [43:20] bag and the shoes. It's safe. That that [43:22] that is the brand promise of a firm and [43:24] it works tremendously well. You know, [43:25] tens of billions of dollars a year [43:26] worth. [43:28] That is now becoming more and more about [43:31] creating a platform for merchants to [43:34] tell their shoppers, hey, we are [43:36] launching a new thing. We have this [43:37] desirable new product we are selling. [43:42] We've transacted with roughly 50 plus [43:44] million Americans alone and we're now [43:45] live in four countries and and growing [43:47] pretty quickly. And so we have shifted [43:50] to just satisfying demand to helping [43:52] merchants create or guarantee demand. [43:54] And the convergence of payments and [43:57] advertising is upon us. Although 15 [43:59] years later predicted it, [44:01] >> you're too early with travel pay. I'm [44:03] certainly too early with PayPal. [44:04] >> Well, the the other side I remember [44:06] maybe it was in that thread or not. I [44:07] should I should find that email. But [44:09] most companies have very high customer [44:12] acquisition cost. And like the joke that [44:13] I make now as a VC is that when I see [44:16] 90% of consumer companies, I'd kind of [44:18] rather buy Google stock or Facebook [44:20] stock because that's where all the [44:21] customers come from. And I distinctly [44:23] remember many many conversations around [44:25] how do we have negative CAC, right? And [44:28] like I think to me one of the coolest [44:30] things about a firm to this day is it [44:32] has negative customer acquisition cost. [44:34] I mean, it is paid to acquire a [44:36] customer. And there there are some [44:37] companies that you would think they look [44:39] like this, but they don't because [44:41] they're basically white labeling [44:42] something. But going back to the [44:44] mattresses, um, let's just say that [44:46] you're late on one of your payments. [44:48] Casper, the friendly white ghost [44:50] mattress company, because Casper is a [44:51] white ghost, doesn't want to go or cat [44:53] or something. We had a cat named Casper [44:55] when I was growing up. They don't want [44:56] to send you a dunning notice saying [44:57] like, "Get off my mattress. You haven't [44:58] paid me back." They want that to be a [45:00] third party versus other companies where [45:04] yes, you know what is the difference? [45:05] What B2B TOC is a really interesting [45:07] category. It's business to business to [45:08] consumer. I tried doing this with trial [45:11] pay because it's like all right, you're [45:12] playing a Zinga game. Go get uh free [45:16] coins if you sign up for Geico powered [45:17] by trial pay. Nobody knew who the hell [45:19] trial pay was. So we we actually tried [45:21] sending messages to consumers like you [45:23] know we got I think we got blocked by [45:25] our email provider because nobody we we [45:28] actually had a right to email these [45:29] customers but they had no idea who we [45:31] were whereas actually having a financial [45:33] relationship with 50 million people with [45:36] negative CAC. I mean it really is not to [45:39] pet but it's incredible like it's so [45:41] hard to do and we never see anything [45:43] like this in the venture land or you [45:45] know like because yeah there are lots of [45:46] companies that can get to scale. I mean [45:49] Casper got to scale but lots of CAC lots [45:51] of CAC and that's why a lot of these [45:53] business models they tend to degrade [45:54] over time because all the economic value [45:57] goes to Google and Facebook so if you [45:58] can have negative CAC and then you can [46:00] launch other products as well right [46:01] because you actually own the customer [46:03] and most important the brand wants you [46:06] owning the customer like this is what I [46:08] didn't get right at trial pit like Zinga [46:10] didn't want me owning their customer [46:12] Netflix didn't want me owning their [46:13] customer I just connected the two but um [46:16] a firm the the merchants She wants you [46:18] to own the customer. [46:20] >> It's very you already have a burden of [46:22] supporting your product and dealing with [46:24] unsatisfied customers or technical [46:26] support issues or whatever. A company [46:29] that tells you 12 times for your long [46:31] loan or 39 times, you know, if it's a [46:34] three and a half year loan, hey, you're [46:37] late or you got to got to make your [46:39] payment or all the communications that [46:41] come from your lender is just another [46:43] type of burden that the brand is kind of [46:45] like, you know what, a firm you guys can [46:46] handle this. This would be great. And so [46:49] the the relationship with the customer [46:52] is a key component of the foundation of [46:56] a firm and the our ability to develop [46:57] new products is is really built on that. [47:00] One of the other really sort of cool to [47:02] sort of going all the way back to the [47:03] credit scoring thing. [47:05] So we are the only company in the [47:07] industry really there's plenty of [47:08] competitors who are trying and some some [47:11] degree of success but mostly not really [47:13] who will go into these longerterm loans [47:15] and by longer I don't mean mortgage for [47:17] you know 15 plus years but like three [47:19] and a half years is a fairly long period [47:21] of time relative to the average bailator [47:24] which is like 6 weeks and to do that you [47:27] have to underwrite like you can't [47:29] shortcut the I'll just look at your FICO [47:31] score or I'll just sort of you know I'll [47:32] look at your Facebook friends like none [47:34] of that works you actually have to do a [47:35] real very sophisticated degree of u [47:38] machine learning work. There are two [47:41] cool things about it. It's hard to do [47:42] and so it's just hard to compete with us [47:44] on that front. But the other side is it [47:47] gives you 39 shots on gold or 12 shots [47:49] on gold to tell this consumer as you [47:51] send them billing notices and as you [47:53] have conversations with them as they pay [47:54] you to upsell them on a new service. And [47:57] so these long-term loans, which are [47:58] actually quite hard to manage and quite [47:59] hard to maintain reasonable default [48:01] rates and and delinquency rates on, is [48:04] the price you pay as a provider of [48:06] financing to upsell consumers on more [48:08] and more services. And so our our [48:10] business has been fueled by these [48:12] long-term products, [48:14] even though they are probably the the [48:16] most challenging part of payments. [48:18] >> Yeah. You know, we've been covering some [48:21] a firm history. While we're doing a [48:22] historical episode, I also want to ask a [48:23] question about PayPal history. [48:26] >> [snorts] [48:27] >> I'll keep us straight [laughter] [48:28] and [48:29] >> you know um there's been you know a book [48:31] came out founders recently there's been [48:32] sort of a lot of uh you know revisiting [48:34] of of of PayPal both in its impact as u [48:38] you know pioneer in sort of uh you know [48:40] payments and and and fintech but also in [48:42] terms of the the people uh that came out [48:44] of it and sort of the it it feels like [48:45] the you know highest concentration of um [48:49] highest hit rate of people who would go [48:50] on to do incredible things. The only [48:53] comparable thing I could think of today [48:54] is it's not even a company is the deal [48:55] fellowship or or something. I can't [48:57] think of a comparable company to PayPal [48:58] today. I'm curious if you you can um and [49:01] you know all with uh or many with [49:03] extremely strong strengths but also you [49:05] know very strong personalities and some [49:07] strong weaknesses as well. And so I'm [49:10] curious if there's something you think [49:11] that is goes as that story is being told [49:13] of both of its impact in the industry, [49:16] you know, how early was uh and and sort [49:19] of the people that came out of it. if [49:20] there's something that's [49:21] underappreciated or you think not quite [49:24] groed or any any additions you want to [49:26] make to how we perceive that that [49:29] company in that time period [49:30] >> sort of an old answer and a new answer [49:32] to this question which I've you might [49:34] you might expect I've answered once or [49:36] twice before u so the the canonical [49:38] answer that I gave in a book which by [49:40] the way is very good so there there have [49:41] been many much ink spilled on the [49:43] history of PayPal the founder is [49:44] actually a very good book like and part [49:45] of it is because it's well written but [49:46] part of it is because Jimmy the guy who [49:48] wrote it interviewed every single [49:49] character in the story extensively. He [49:51] spent a lot of time talking to all of us [49:53] and so it's at the very least grounded [49:55] in fact. Um so the the answer I gave for [49:59] his research was as we were interviewing [50:02] certainly the early team and even the [50:03] later team we kept on selecting four [50:05] entrepreneurs. We would ask like one of [50:07] the stock questions I would ask so what [50:08] are you going to do after PayPal? And [50:11] the answer that I welcomed the most was [50:12] oh this is the last one I'm going to [50:14] start my own afterwards. So this notion [50:15] like oh if you're coming in here to kind [50:17] of get your final graduate degree in [50:20] entrepreneurship of course we'd love to [50:21] have you like you know if you're you're [50:23] that smart you're that good and you're [50:24] that ambitious you know please come [50:25] along and so that's kind of the why I [50:29] think so many people that came out of [50:31] PayPal proceeded to start companies and [50:33] like literally you know hot second after [50:36] we all dispersed YouTube and Yelp and [50:40] Peter's first major fund and LinkedIn [50:42] and so that was not an accident we for [50:44] attracting entrepreneurs by design. The [50:47] other part that I think is actually not [50:49] been talked about and maybe subtle and [50:50] sort of my own cut of this, but we knew [50:53] the people in the room intimately like [50:55] we literally, you know, exchanged firm [50:57] by hanging out in these sweaty rooms, [50:58] brainstorming, screaming at each other [51:00] and spending a lot of time as [51:01] whiteboards and declaring the other [51:03] person's ideas idiotic and and so on. we [51:06] knew the the true version of that [51:08] person, the sort of the extremely [51:12] I don't know uh the the the ultra [51:14] philosophical Peter Teal and the uh you [51:17] know go all in every time Elon Musk and [51:20] the [51:22] these days uh gubernatorial uh David [51:25] Sax, you know, all these people like [51:26] they all have kind of the the present [51:28] the presentation layer and then there's [51:29] like the kind of what they are like in a [51:32] dinner party and then like what are they [51:34] when they're really stressed out and [51:36] just like head hurts, don't know what to [51:38] do, and we have a real problem we need [51:39] to solve. So, if you know kind of the [51:41] true base version of who we all are as [51:46] you watch one of these people go for [51:48] something really big like Elon taking [51:51] humanity to Mars, which is [51:53] extraordinary. And yet, you know, he's a [51:56] human being. Like, I've seen him sweaty [51:58] and tired and grumpy at the company [52:00] kitchen a thousand times. or Peter when [52:04] he is in his doubt and calls me from [52:05] some fundraising trip saying, "I think [52:07] we might run out of money." You know, [52:09] they're not gods. You know, they're [52:10] humans. And that inspires you to [52:12] actually like, you know what, if that [52:14] guy who's just a normal dude, I spent a [52:15] lot of time with him, he's normal. [52:18] And he's going for this big of an idea, [52:21] I should do the same. I I should I [52:23] should strive, too. And so, I think the [52:24] combination of these are very ambitious [52:27] people. We all wanted to be [52:27] entrepreneurs, but we also knew each [52:30] other when we were young and filled with [52:32] doubt, which many I think still are. Uh [52:34] I think that that sort of gives you a [52:36] confidence that it's not all, you know, [52:39] bed of roses. It's it's it's paved with [52:41] uh with with with with bumpy parts. [52:44] >> Is there anything we didn't get to that [52:45] you want to make sure that we get to [52:47] while we're here? [52:48] >> I'm probably less optimistic about [52:51] agentic shopping and I'm very optimistic [52:54] about aentic payments. Yeah, [52:57] >> I think the the notion of robots will [53:00] buy our [53:02] Friday night outfits is misguided. We [53:04] want to know what we look like long [53:06] before the robot delivers it to your [53:07] door. But I think the friction, this [53:09] notion of so the best user interface [53:11] ever created is the credit card. This [53:14] may actually be finally up for [53:16] renegotiation because agents are in fact [53:18] smarter than pieces of plastic and even [53:20] pieces of plastic with rewritable chips. [53:22] And so I think that's uh a lot of [53:25] conversation around aentic commerce is [53:27] probably looking 20 degrees off where [53:29] you know how can I convince you to just [53:31] trust a robot to buy the right thing. I [53:34] don't want to like I love buying bike [53:35] parts and I spend way too much time [53:36] looking at two slightly different bike [53:39] parts because I just you know have to [53:41] participate in the purchase [sighs] [53:43] but [snorts] the part where like okay [53:44] that one the end right now that one is [53:48] the beginning where you pull out your [53:49] wallet or you and so I think we we're [53:52] we're about to see some interesting [53:54] innovation in payments user interface. [53:56] >> So I can I ask a question on that? So, [53:59] um I mostly agree, but I think it [54:01] depends on there's like the I need to [54:03] research what to buy. [54:04] >> Mh. [54:05] >> Where AI is just yet another tool at [54:07] your disposal. Like you ask your [54:08] friends, that's what I'm if it's a bike [54:10] part, I'm asking you. I'm also asking [54:12] AI. I'm asking both of you. But now I [54:14] have the skew that I want. And I don't [54:18] actually care which one of the 19 [54:19] different places that I get it from. And [54:21] I kind of think about the world like you [54:23] have people that care more about money [54:24] than time. You have people that care [54:26] more about time than money. [54:28] The way that I kind of think about [54:29] ajected commerce, again, none of this is [54:31] happening right now. So, it's trying to [54:32] prognosticate what's going to happen is [54:35] here is the the UPC. Here's the the skew [54:37] that I actually want. Go buy it for me [54:39] at the lowest cost. It's kind of the [54:42] thing that people that have a lot more [54:43] time than money do right now anyway. And [54:46] like that's why I always point to Camel. [54:47] Camel. Camel is like my favorite example [54:49] of this because it's one of the top [54:52] hundred websites in America that [54:54] probably nobody with over $100,000 a [54:56] year of income has ever heard of. But [54:58] it's it's a way of saying I know what I [55:00] want. You're not helping. The robot's [55:02] not telling me what bike part to buy, [55:03] what telling me what soda to buy. I [55:05] know. Just give it to me at the lowest [55:07] price. But it takes way too much time. [55:10] And sometimes you can just observe one [55:12] group of humans that already have [55:14] something as table stakes and then [55:16] porting that to another group of humans. [55:17] I I guess do you think that that form of [55:20] agentic commerce because it's actually [55:22] related to edgentic payments as well. [55:23] >> I have nine different cards. I don't [55:26] remember the Amazon Chase issued by [55:29] Chase card terms for a nonrest purchase [55:32] of something under $500. and if I've [55:34] used my whatever. Like it's the same [55:36] thing. Like people that care more about [55:38] money than time, they will figure that [55:40] out. People that care more about time [55:41] than money, they probably won't. I think [55:44] directionally that'll make sense. I [55:47] think the adoption curve is probably [55:52] going to be fairly slow. And these [55:54] things are always sort of easy to [55:55] predict to be very slow. And sort of the [55:56] usual adage is well slowly at first, not [55:58] always once. I'm I'm not I'm not sure I [56:01] I can handicap that. Well, the reason [56:05] people don't So, there's plenty of [56:07] people who live in this somewhat murky [56:10] world of sometimes I have money, [56:11] sometimes I have time. And it has more [56:13] to do than anything with the price and [56:15] the degree of care you put into. I mean, [56:17] I don't always just buy bike parts. [56:19] Sometimes I buy milk and milk is built. [56:21] Where are my bike parts? [56:24] I mostly buy parts. Um, but uh the I I [56:28] think what really happens for a lot of [56:31] people, it's just like, well, yeah, I'll [56:34] I'll get the I'll find the cheapest [56:36] price, but then I'm not sure how to [56:38] handicap the probability of that item [56:41] actually showing up at my door at the [56:44] time of my desired outcome. Even if I'm [56:47] okay buying a some bike part from some [56:50] far away land and it's sold from three [56:52] different retailers, [56:54] I'm going to go with the most reputable [56:56] one versus the one that's got profit [56:58] promising me the cheapest or sometimes [56:59] the fastest. And so I think we're [57:02] probably a few steps away from AI [57:05] grasping [57:07] both the preferences but also [57:08] handicapping how these preferences would [57:10] change based on the input which may in [57:14] fact maybe AI is already there. is just [57:16] that you haven't yet trusted your agent [57:18] to do as good a job as you would. Like [57:20] you can eyeball a side be like that [57:21] looks like it was edited in the 90s. I'm [57:24] just not quite sure that I'm going to [57:25] spend $1,000 for a new cassette or $600, [57:28] whatever the cassettes go for these days [57:30] in bike cassettes, the rear assembly. [57:33] >> Well, Bike Closet is the site that I'm [57:34] thinking about that looks like it was [57:35] made in the 1990s that will sell you a [57:37] very cheap cassette. [57:38] >> Uh, exactly. But do you know if they [57:40] were used by by someone you know true [57:43] teeth rubbed off and all? [57:45] >> Uh yeah. So I I think I think we are in [57:48] that difficult period where we're still [57:51] figuring out. The other thing that you [57:52] know just sort of maybe end on a [57:54] positive note. The thing that's really [57:56] true that no one seems to be celebrating [57:59] just yet is [58:01] grocery shopping is 100% agentic. Like [58:03] for anyone who uses Instacart like you [58:05] tell your Instacart shopper go bring me [58:07] milk. you don't think twice when they [58:09] say, "Oh, yeah, you wanted whole milk [58:11] from Organic Valley, but I found some [58:15] other brand and here it is." And you're [58:17] like, "Oh, of course, like makes sense." [58:18] Like, and so the AI inside that [58:20] shopper's head is maybe not even as good [58:22] as Gemini or ChadB is these days, but [58:24] you just sort of say, "I you just you go [58:26] through Trader Joe's or Whole Foods or [58:28] whatever you're shopping and you figure [58:29] it out and it shows up and 99.9% of the [58:32] time it is exactly what you want it or [58:34] better." And so we are already [58:35] conditioned to allow some of these [58:38] purchases to be fully outsourced, [58:40] payments and shopping and everything [58:42] else. And so we're definitely going to [58:43] get there. It's just a question of how [58:45] soon will we work through the quirks of [58:47] like well returning of this wrongly [58:50] purchased bike part or waiting for it [58:52] forever and not really knowing what's [58:54] going to happen to it because it's in [58:55] transit but we don't know where it is. I [58:57] that may be protracted more protracted [59:00] people think. [59:01] >> That's a good note to end on. We'll have [59:03] to uh get a part two uh at some point [59:05] when we talk more about the future. Uh [59:06] Max, Alex, thanks so much for