Guo Bets Conviction Against Big Tech Monopoly in Frontier AI

郭押注坚定信念 对抗前沿AI的大科技垄断 Sarah Guo认为，开源模型与独立资本对于防止少数前沿实验室控制整个AI经济至关重要。 Sarah Guo阐述了其投资机构在缺乏历史回测的空前技术浪潮中，支持早期AI原生公司的投资逻辑。她探讨了前沿研究人员之间激烈的人才争夺战与心理转变、算力基础设施瓶颈，以及她为何支持开源模型而非单一巨头实验室的垄断。郭还分享了法律科技、机器人和生物医药领域的详细案例，强调了基于第一性原理的技术评估、具备高能动性（high-agency）的创始人以及注重执行的坚定信念，如何在硬科技与非传统软件市场中带来超额回报。 当前AI时代缺乏历史回测，使风险投资人在踩刹车与高速推进之间产生了强烈的张力。 (2:16) 在风险资本支持下，具备高度能动性的个体能够影响宏观AI走向，防止少数单一的前沿实验室控制整个生态。 (3:31) 对Harvey的早期投资表明，发掘AI机遇的一种非显见方法是将客户需求与对模型能力的前瞻性技术评估直接结合。 (8:37) 前沿AI研究人员正面临存在危机，感到个人的人类智慧正日益被海量算力扩展和资本投入所遮蔽。 (11:00) 算力可用性与物理基础设施——尤其是基础电力供应和现实世界的供应链限制——是AI取得进展的主要瓶颈。 (13:50)

Sarah Guo, founder of venture firm Conviction, argues that navigating the current AI transition requires backing high-agency individuals and pairing deep technical intuition with execution speed, rather than relying on traditional historical backtests or uncritical capital. Navigating Generational AI Without Historical Backtests (2:16) Guo describes early-stage AI investing as operating under an intense dilemma: managing the risk of missing a major technological shift while avoiding capital destruction in a potential boom-and-bust cycle. Because traditional venture backtests fail in this rapid transition, investors often feel the conflicting urge to apply the brakes while moving at 90 miles per hour. Guo rejects the fatalistic assumption that a handful of closed frontier model labs will monopolize the economic gains, citing the role of high-agency founders and open-source alternatives in shaping broad technological diffusion. Technical Premises and Novel Application Archetypes (8:37) Rather than relying strictly on conventional customer-back frameworks, Conviction combines customer needs with technical assessments of underlying model capabilities. In backing Harvey in late 2022, Guo and her team recognized that the structured nature of legal text aligned directly with large language models' next-token prediction strengths, progressing from routine contract analysis to automating large portions of major M&A transactions. Similar dynamics appear in biology: after years of computational biology failing to capture software margins, foundation models enabled Chai Discovery to secure eight-figure software and platform contracts directly with major pharmaceutical enterprises. Physical Constraints, Energy, and Compute Independence (13:50) Guo highlights compute availability and real-world infrastructure as primary bottlenecks holding back artificial intelligence. Citing a conversation with a hyperscaler infrastructure leader, she notes that new energy projects are unlikely to meaningfully move the needle at scale before 2030 due to regulatory hurdles and supply chain limits around natural gas and small modular reactors. To prevent industrial paralysis, Guo advocates for 'compute independence,' noting that sovereign and economic resilience requires securing upstream dependencies—such as specialized glass manufacturing for semiconductor foundries—alongside investments in hardware, nuclear energy, and robotics. Empirical Robotics and High-Variance Data Collection (20:00) In robotics, Guo highlights founders Tony Chen and Zipeng Fu of Physical Intelligence, who approached the historical scarcity of robotics data as an engineering problem rather than a capital-intensive brute-force challenge. By developing low-cost methods to capture rich distributions of real-world environments and transferring those datasets into model training, the company condensed development timelines from Stanford research prototypes to deploying semi-humanoids for domestic beta testing within two years, shifting the industry consensus on generalized robotics from a question of possibility to a near-term reality. Decision Mechanics and Partner-Level Conviction (22:15) Conviction's internal process relies on identifying strong initial intuition about a founder's industrial logic and then working backward to aggressively stress-test assumptions through written memos and external peer reviews. Guo rejects consensus-based committee decisions and uncritical funding rounds driven purely by social proof or pedigree. Looking ahead, she anticipates that the Jevons paradox will apply across non-engineering knowledge work: as AI tools automate tasks in functions like marketing and sales, organizations will raise their baseline output goals and resource consumption rather than reducing overall working hours. Guo maintains that venture returns in the AI era will stem from execution speed and primary interactions with researchers rather than speculative secondary commentary, though significant physical constraints in power and supply chains remain unresolved bottlenecks for frontier model scaling.

Conviction 创始人 Sarah Guo 认为，当前人工智能浪潮缺乏历史回测经验，投资者不应盲目押注少数垄断性前沿实验室，而应通过第一性原理和高能动性人才，将资本聚焦于底层算力瓶颈突破以及向实体经济深度渗透的 AI 原生应用。 无回测周期的投资困境与能动性理论 (2:16) 当前 AI 投资生态面临着极度焦虑的节奏失衡，投资人普遍在踩死刹车的冲动与时速 90 英里的狂奔之间拉扯。郭指出，历史上的传统风险投资回测模型在当下颠覆性的技术转移中已经失效，基金既要防范错失世代机遇，又要规避科技繁荣与萧条周期带来的资本毁灭风险。针对前沿实验室可能垄断全部经济价值的悲观预期，她援引“伟大人物历史理论”予以反驳，强调具备高能动性的个体在获得匹配的风险资本后，能够创造出抗衡垄断的替代方案（如开源模型），从而根本性地重塑产业格局。 技术前置评估与非共识应用挖掘 (8:37) 与传统仅从客户需求出发的框架不同，Conviction 采取将客户痛点与底层模型演进能力相结合的双向评估机制。以法律 AI 公司 Harvey 为例，团队在 2022 年底洞察到大语言模型的下文预测机制与高度结构化、重检索的法律文本天然契合，该项目随后从基础的租约审查演进至能够自动化处理动视暴雪并购案中高达 85% 的法律工作。然而，郭也坦承认知局限带来的失误，例如曾因低估大众的音乐创作欲望而错失 Suno，这促使她更加重视一线研究人员和真实用户行为反馈，而非沉溺于应用层是否具备防御性的二次舆论争辩。 算力瓶颈与物理供应链重塑 (13:50) 郭强调低成本智能绝非必然，算力供给是制约 AI 发展的最大现实瓶颈。某大型云厂商基础设施主管向其透露，在 2030 年之前几乎没有新的大型能源项目能实质性缓解供电压力，小型模块化反应堆（SMR）与天然气基荷电力的落地主要受制于监管审批而非技术本身。她提出“算力独立”概念，指出供应链不仅依赖先进晶圆代工，更受制于台积电上游特种玻璃单一供应商等脆弱环节。巨大的供需失衡促使 Conviction 颠覆过去视芯片为高危投资的传统信条，积极评估硬件、核能及数据中心机器人等硬科技瓶颈。 前沿模型突破：从具身智能到生物医药 (20:00) 在机器人领域，Physical Intelligence 的核心创新在于摒弃耗资数亿美元的暴力数据采集模式，转而通过低成本、高方差的环境数据生成方法打破“数据荒”，并在不到两年内实现软硬件一体化迭代，计划年内展开半人型机器人家庭公测。在生物医药赛道，Conviction 作为首家机构投资了 Chai Discovery，该团队通过大模型协助全球排名前十的药企加速早期研发，并达成了高达千万美元级别的纯软件与平台合作协议，打破了以往计算生物公司必须转型为承担临床管线风险的传统商业模式局限。 逆向确信机制与杰文斯悖论的蔓延 (22:15) 在投资决策机制上，郭倾向于在初次接触项目时迅速建立 8 到 9 分的直觉判断，随后通过撰写详尽的投资备忘录并交由外部同行严苛推敲，以此逆向寻找认知盲区，且要求合伙人个人对投资结果承担明确兜底责任。展望未来一年，郭预测“杰文斯悖论”将在非工程类知识工作中全面显现：正如 AI 辅助编程提升了开发者产能而非减少岗位，单个营销人员构建的“自主营销部门”大幅提高了交付物总量，效率工具的普及将刺激更高密度的产出需求，而非缩减整体工作负荷。 郭确信，硅谷最核心的优势在于忽视履历标签、押注人类潜力的信任文化。但在算力基建受限于物理电网与地缘供应链的严峻现实下，极速迭代的模型能力能否在 2030 年前顺利完成向更广泛实体经济的普惠渗透，仍是技术与资本市场悬而未决的关键考验。

## Background 本期节目由 Colossus 旗下的播客《Invest Like the Best》于 2026 年 9 月 1 日在 RSS 及各大播客平台发布。节目主要围绕早期风险投资机构 Conviction 的创始人兼管理合伙人 Sarah Guo 展开，深入探讨了当下 AI 前沿突破者的核心信念、顶尖研究人员的心理困境、具身智能与科研加速的进展，以及为何未来绝不会被单一模型巨头垄断。 ## TL;DR Sarah Guo 在对话中系统阐述了 Conviction 聚焦 AI 前沿的投资逻辑，主张反抗闭源实验室的寡头垄断，全力下注由高能动性人才与开源生态支撑的多元未来。她揭示了当顶级科研被万亿算力竞赛绑架时，核心研究人员所面临的自我价值虚无感，并重点剖析了具身智能低成本数据采集、生物计算模型驱动重构，以及国家工业安全维度的“算力独立性”瓶颈。Sarah 还回顾了个人创办机构的经历与决策机制，强调早期投资的精髓在于面对不确定性时主动搁置怀疑、推行合伙人全权担责，并在现实证据完全显露前，凭借第一性原理与直觉坚定押注真正优秀的人才。 ## Atmosphere & style 整场对话在一种快节奏的技术紧迫感与冷静的智识反思之间保持着微妙平衡。主持人的提问直击当下 AI 狂热背后的盲区与泡沫焦虑，而 Sarah Guo 则以极其严密的第一性原理逻辑、坦率而不迎合的语言风格，将宏大叙事层层拆解为微观的工程与商业现实。对话既有探讨国家基建与万亿资本开支时的肃穆厚重，也有聊及具身智能突破、父母创业教诲以及硅谷纯粹信任时的真挚与热忱，给听众带来一种身处历史转折点火线前沿、既亢奋又清醒的沉浸式智力体验。 ## Bulletin - `06:16` 全球仅约 250 位核心研究者与架构师在真正推动 AI 技术前沿，早期投资的核心策略在于全面渗透这一核心网络，而非盲目海选。 - `10:00` 随着模型递归自我改进和千亿美元算力扩张，前沿科研面临组织工程化膨胀，部分顶尖研究者正陷入个体贡献被边缘化的意义危机。 - `13:50` 算力与电力基础设施的物理供应链瓶颈难以在 2030 年前靠纯技术手段逆转，算力独立性与供应链冗余成为国家与产业竞争的核心命题。 - `34:38` 开源 AI 模型在成本、延迟与隐私约束下已呈现不可逆的扩散趋势，依靠闭源 API 无法覆盖实体经济中 99% 的复杂落地场景。 - `45:27` 生物计算正在打破传统药企软件无法变现的铁律，底层基础模型已能直接切入大药企核心研发管线并获取千万美元级商业大单。 ## Key takeaways - `00:00` 历史回测在当前 AI 技术质变期全面失效，早期投资人深处既害怕错失时代机遇又恐惧重蹈泡沫覆辙的紧绷状态。 - `03:31` Conviction 坚决反对极少数闭源前沿实验室垄断经济全产业链，主张通过资本赋能高能动性人才与开源生态来维系去中心化格局。 - `06:16` 推动人类 AI 前沿的核心人才全球仅约 250 人，基金的超额认知壁垒源于无限贴近这一极小圈子并理解其技术追求。 - `08:37` 非共识投资需绕开传统从客户痛点倒推的 SaaS 框架，转而寻找底层技术本质能力与垂直数据结构的天然契合点（如法律的结构化自然语言与大模型预测）。 - `10:00` 算力规模化与模型自我迭代引发顶尖学者心理异化：研究者面临个体无法改变战局或一切努力皆被算力取代的虚无感。 - `13:50` 物理供应链、熟练工人储备、核能基荷电力与监管审批形成算力供给硬约束，2030 年前无法靠纯软件迭代方式化解。 - `13:50` 资本市场在面对前沿技术时出现过度依赖毕业院校和名企履历等代理信号的非理性繁荣，忽视了商业模式与技术可行性的独立直觉。 - `20:00` 具身智能的突破口在于以极低成本巧妙采集物理交互数据并高效转移到模型训练，家庭通用机器人迭代速度远超业界原本预期。 - `22:15` 投资决策应建立在高度敏锐的直觉之上，但必须辅以详尽的备忘录和反向证伪流程，主动寻找认知盲区与反例以校准判断。 - `26:13` 时间精力需严格聚焦：60% 投入投后深度赋能，各 20% 分配给高信噪比的项目精选及跨资产类别的深度认知拓展。 - `30:49` 反感创投圈 Series A/B 阶段零和博弈与做秀文化，推崇像公开市场般的正和分享，以及创业者对抗垄断巨头的产品硬实力。 - `34:38` 开源 AI 模型是实现技术平权与广泛商业渗透的基石，单靠行政禁令只会惩罚守法企业，无法阻挡全球技术扩散。 - `39:01` 丰沛廉价的智能并非历史必然，美国本土制造业断代亟需依靠高度自动化与充足低成本算力维系国家工业竞争力。 - `40:00` 算力独立性关乎国家安全，核心在于消除单点瓶颈并构建供应链冗余，而非全产业链闭门自造。 - `43:14` 半导体创投的传统赔钱铁律被超级买家摆脱代工依赖的诉求所打破，早期风投辩论的核心转向市场能否容纳幂律回报分布。 - `45:27` Chai Discovery 等案例表明基础模型已能打破生物医药只认管线不认软件的旧范式，以核心研发增益直接换取大药企商业价值。 - `48:58` 早期投资的灵魂在于主动搁置怀疑并实行合伙人个人全责制，拒绝稀释责任的委员会决策，以坚定信念陪伴企业穿越迷雾。 - `50:00` 投资的关键是探寻被错误定价的真实，避免陷入大模型巨头战略博弈的循环口水战，重注 AI 在真实业务中扩散的剩余 99%。 - `54:16` 杰文斯悖论将在 AI 落地中全面显现：平庸工作的边际成本下降将释放组织生产力，催生出更庞大的业务需求与更高的劳动投入。 - `56:46` 硅谷最珍贵的生态特质在于抛开资历偏见、在缺乏确凿证据时纯粹基于潜质与思想给予信任，这正是投资与创业的核心信仰。 ## Detailed walkthrough ### `00:00` 狂热周期中的投资困境：失去“回测”的历史时刻 对话在一种弥漫整个 AI 投资界的紧迫与狂热氛围中展开。Patrick O'Shaughnessy 指出，当前身处技术浪潮中的人无一不感到忙乱和紧绷，而且这种混沌感不仅没有放缓的迹象，反而可能愈演愈烈。对早期投资人而言，最核心的焦虑在于：面对一场与以往截然不同的技术质变，所有传统的“历史回测”（backtest）与既往经验似乎都失效了。 Sarah 提到前一天晚上与一位投资人朋友的交流，对方给出了一个极度形象的隐喻：“我总跟自己说必须竭尽全力踩死刹车，可现实中我却开着车以 90 英里的时速在狂飙，根本停不下来。”这精准地折射出当下投资人的双重恐惧：一方面害怕因犹豫而错失划时代的历史机遇；另一方面又恐惧重蹈科技史上每一次“繁荣与萧条”（boom-bust）周期的泡沫覆辙。尤其对于 Conviction 这样刚成立数年的新兴机构来说，如何在如此高波动的周期中平衡机构的基业长青、团队的职业发展与高风险投资决策，是一个极其复杂的命题。 ### `03:31` AI 的终局赌局：反对前沿巨头的寡头垄断 Patrick 提及 Colossus 团队此前为 Sarah 撰写的特写报道《Sarah's Wager》（莎拉的赌注），文中将她的投资策略描述为一种鲜明的抗衡立场——在极少数几家底层模型实验室（如 OpenAI、Anthropic 等）与广泛的外部生态之间，她选择重注后者，以防技术终局走向赢家通吃的单极世界。 针对这种“宣战”式的解读，Sarah 从“英雄史观”（Great man / great woman theory）的角度给出了更深层的剖析： - **高能动性个体的力量**：历史并非由机械的宿命决定。当极少数拥有极致能动性（high-agency）的人才获得充沛的风险资本和生态环境支持时，他们确实有能力扭转历史走向。 - **开源生态与产业安全**：以开源模型和美国产业政策为例，未来是否能诞生一个具备全球竞争力的西方开源大模型生态，完全取决于现实中是否有人敢于站出来，承担巨额资本开支、召集最顶尖的人才底座、搭建算力基础设施，去在前沿高地上真刀真枪地拼杀。 - **拒绝寡头垄断的未来**：虽然她与各大前沿实验室保持着密切合作、共同投资且私交甚笃，但她坚决反对一种极端论调——即一两家拥有顶级前沿模型的巨头将吞噬整个经济体。“我非常明确地不希望看到那样的未来，而且我认为我们绝不会落入那样的境地。” - **投资人的自我定位**：当被问及是否渴望成为改写历史的“伟大女性”时，Sarah 坦承自己毫无这种个人英雄主义的虚荣心。她更想成为所处领域里最卓越的投资人，本质是由极度旺盛的好奇心、对“把事情做对”（being right）的执念以及同顶尖人才并肩作战的渴望所驱动。她选择退后一步，通过资本与技能去成就那些真正改写世界的创业者。 ```mermaid flowchart TD accTitle: Conviction 投资理念与生态平衡 accDescr: 展示 Conviction 如何通过资本赋能高能动性人才，抗衡闭源寡头垄断并构建多元 AI 生态。 subgraph Labs["闭源前沿实验室 (Frontier Labs)"] L1["极少数闭源巨头"] L2["试图吞噬全产业链经济价值"] end subgraph Agency["高能动性个体与生态 (High-Agency Builders)"] T["顶尖前沿人才 (约250位核心圈)"] C["重度风险资本 & 基础设施"] OS["西方开源模型 & 垂直应用"] end subgraph Vision["未来图景"] V1["打破单极垄断格局"] V2["多元、健康的去中心化 AI 经济体"] end L1 -.->|"产生垄断威胁"| L2 T -->|"注入资本与支持"| C C -->|"构建"| OS OS -->|"对抗与制衡"| L2 OS -->|"最终达成"| V1 V1 --> V2 ``` ### `06:16` 创立顶尖投资机构：250 人的前沿核心圈 对于 Conviction 为何能在短时间内脱颖而出，甚至在各大顶级机构有限合伙人（LP）的匿名偏好调查中名列前茅，外界常常感到好奇。Sarah 认为前期的战略逻辑其实朴素而纯粹：传统老牌风险投资机构伴随规模膨胀正在经历漫长的代际更替，早期投资领域的敏锐度出现断层；而此时正值巨大的技术范式转移，只要敢于承担风险、从第一性原理出发深扎技术与开发者社区，就能建立起传统综合型基金无法企及的认知深度与项目触达能力。 然而，Patrick 指出单纯靠“比别人更勤奋”不足以解释其亮眼业绩。Sarah 与其合伙人 Mike 建立的核心认知框架在于： - **前沿人才的极致集中**：在整个人类 AI 前沿技术突破的火线阵地上，真正在清晨醒来、用一行行代码和一次次实验把整部技术列车向前推动的核心人物，全球大概只有 250 个人左右。他们是前沿科学家、顶级架构师与极高天花板创业者的交集。 - **无限贴近核心人才**：Conviction 的核心使命并非盲目进行上百场漫无目的的早期项目初筛，而是全面渗透进这 250 人的核心网络，深刻理解他们的技术追求，并在他们萌生创业构想的第一时间提供一切资源支持。 ### `08:37` 非共识的下注方法论：以技术原生匹配重构软件 当谈及如何发掘那些尚未形成市场共识的早期机会时，Sarah 指出，绝大多数传统投资人依然机械地套用“从客户痛点出发倒推”（customer backwards）的传统软件投资逻辑，并斥责技术驱动论是无稽之谈。但在范式转换初期，Conviction 坚持采取一种极为前沿的“技术与场景强匹配”（technology-forward fit-for-purpose）的选品逻辑。 他们重新审视底层技术能力的本质增长与具体应用领域的契合度。Sarah 详细拆解了他们在 2022 年底决定重注法律 AI 平台 Harvey 时的第一性原理推导： - **技术本质的映射**：大语言模型的核心是海量文本语境下的“下一个 Token 预测”（next token prediction）与概率生成，辅以检索增强技术。 - **法律业务的天然契合**：法律从本质上讲就是高度结构化的自然语言。律师的日常工作涉及海量严密文档的阅读分析、先例判例的检索回溯、以及高度专业化文本的逻辑撰写。无论在律所内部知识库还是英美法系的判例法体系中，历史文本沉淀极为丰厚，这与大模型的技术优势形成了天衣无缝的交集。 正是这种绕开传统 SaaS 评估体系、直接透视模型技术底层与人类垂直行业数据结构匹配度的非共识框架，构成了 Conviction 在浪潮之巅建立壁垒的关键底牌。 ### `10:00` 前沿研究者的心理困境：算力膨胀与自我意义的丧失 在回顾早期支持法律 AI 公司 Harvey 的决策时，Sarah Guo 提到他们对技术潜力和落地价值有着极其清晰的判断。联合创始人 Winston Weinberg 与 Gabriel Pereyra 坚信 AI 会以彻底被“AI 药丸”（AI-pilled）洗脑的姿态重塑法律行业的实践。他们的愿景并不是停留在“帮加州租户阅读租赁合同并回答问题”这种琐碎任务，而是大胆跨越到“主导动视暴雪（Activision Blizzard）这样庞大而复杂的并购交易，并自动完成 85% 的工作”。打动 Sarah 的正是这种在当时极具野心且具备底层技术逻辑支撑的飞跃，这与其他机构的决策框架截然不同。 随后话题转向如今最核心的 250 位前沿研究者圈子，整个竞争格局已经演变到近乎残酷、暴力的程度（violently competitive），且是一场全球化的军备竞赛。在过去 12 个月里，研究者内部开始蔓延一种既充满野心又极度不安的情绪： ```mermaid flowchart TD accTitle: 前沿 AI 研究者的心理转变 accDescr: 展示了算力规模化与模型自我迭代如何导致顶级研究者产生自我价值缺失的心理链条。 subgraph Belief ["新兴信念与范式"] A["递归自我改进<br>(Recursive Self-Improvement)"] --> B["指数级智能<br>仅需 1-2 年即可降临"] C["算力扩张为绝对核心<br>(千亿美元级 CapEx)"] --> B end subgraph Psychology ["研究者心理困境"] B --> D{"个体价值虚无感"} D --> E["模型自身将解决科研，<br>研究者工作变得无足轻重"] D --> F["决定胜负的只有算力规模，<br>个体不再能够扭转乾坤"] end ``` 这种信念认为，随着模型递归自我迭代（Recursive Self-Improvement），人类距离某种“指数级智能”可能仅剩一两年时间。知名学者 Andre Karpathy 曾自嘲“自己觉得距离 AGI 还有两年这句话已经说了十年”，如今他又再次这么认为。然而，这种剧变带来了深刻的心理异化： - **组织膨胀与个体边缘化**：当 OpenAI 只有 200 人时，顶尖科学家能切实体会到“我能凭借一己之力改变战局（move the needle）”；但当项目演变成需要 7500 亿美元算力、数千名员工协作的超级系统工程时，个人对最终成果的掌控感荡然无存。 - **两种消极的心理预设**：许多顶尖学者开始陷入两难——要么认为“我自己做什么都不重要，因为模型很快会代替我去做”，要么认为“除了砸算力堆规模，其他一切努力都是徒劳”。对于渴望做出真正重要贡献的顶尖天才而言，这两种假设都极具剥夺感。 - **投资人的反思**：面对创业者关于“如果没有你注资，这些项目是否还能拿到钱”的提问，Sarah 坦承最多只有 5% 到 10% 的公司会拿不到钱；真正稀缺的顶尖人才极具资源整合能力，但他们后续能否获得上亿美元的算力支撑则是另一回事。 --- ### `13:50` 物理供应链瓶颈与投资界的“背景代理”陷阱 谈到阻碍未来愿景落地的最大隐患，Sarah 毫不犹豫地指向了算力与底层基础设施。行业目光甚至已经投射到了 2032 年的大规模供电与部署。一位头部超大规模云厂商（Hyperscaler）的基础设施负责人曾对她坦言：在 2030 年之前，没有任何现有技术手段能够真正扭转这种规模级算力供不应求的局面。 这一瓶颈的本质并非资金、资本主义体制或工程能力的匮乏，而在于现实世界的**监管障碍与社会共识对齐**： - **社会共识与审批拖延**：若想在纽约建立大型数据中心，难点在于如何让当地公众认同并在行政程序上批准； - **核能基荷电力**：若要让小型模块化反应堆（SMR）的度电成本下降，必须先让公众相信其安全性并允许批量建设，从而在理论走向上压低成本曲线； - **难以逾越的物理供应链**：隐性知识积累（tacit knowledge）、熟练产业工人、天然气获取和关键原材料的开采加工，其发展节奏根本无法匹配软件行业敏捷迭代的脚步，除了硬着头皮持续砸钱投资，没有任何捷径可走。 在资本端，Sarah 观察到了令人忧虑的非理性繁荣。由于绝大多数投资人和资金掌管者并非一线科学家，面对前沿研究型创业者时，许多机构开始放弃对技术底层的扎实研判，转而依赖**可读性更强的代理信号（Proxying Judgment）**： - **唯背景论（Pedigree Proxy）**：看重创业者的毕业院校、知名实验室履历、曾被谁投资过或是业内谁递交的推荐； - **缺乏商业与技术直觉**：Sarah 曾与一位极其优秀的投资人朋友争论一个项目，当被追问“这家公司到底要做成什么大业务”时，对方唯一的答复竟是“你难道不知道这个创始人的履历有多顶级吗”。Sarah 强调，虽然履历是长期的优质信号，但在缺乏对商业模式和技术可行性独立直觉的情况下，盲目豪赌大规模科研项目极具危险性。 --- ### `18:00` 英雄史观与 Sunday Robotics 的天才学者 当被问及是否有令人惊叹的顶尖研究者时，Sarah 坚定地赞同“英雄史观（Great Person Theory）”——历史长河中的确存在少数关键人物，他们凭借一己之力打破平庸的进展轨迹，在技术曲线上划出一道陡峭的拐弯点（kink）。 Sarah 与她的合伙人 Pranav 挖掘并投资了 Sunday Robotics 的联合创始人 Tony Zhao 与 Cheng Chi。他们此前曾是斯坦福大学的年轻博士生（当时大约仅 25 岁，其中一位甚至为了创业提前退学），但已经拥有在丰田研究院（Toyota Research Institute）、DeepMind 以及特斯拉（Tesla）等顶尖机构深耕的深厚实践积淀，展现出极为超常的技术敏锐度和工程执行力。 ### `20:00` 具身智能与低成本数据采集的突破 在谈及具身智能与机器人领域时，Sarah Guo 重点回顾了她投资由 Tony Zhao 和 Zipeng Fu（傅子鹏）等人创立的具身智能机器人公司（Physical Intelligence / 类似前沿团队）时的震撼感受。她指出，这两位极度年轻的研究者几乎凭一己之力，在过去四年里贡献了机器人 AI 领域绝大多数最有趣的原创思想。在传统机器人学界，大家普遍持有这样一种信念：如果机器人领域拥有如同互联网一般海量的数据，通用机器人早就遍地开花了。但现实中最致命的瓶颈在于——这些物理交互数据从何而来？ 这两位创始人没有坐等无限的算力或资金去堆砌昂贵的数据集，而是展现出极高的工程与科研创造力：思考如何以最低廉的成本，设计极其巧妙的数据采集方式，以覆盖现实世界多样化的环境与任务分布。他们将数据短缺与资金约束视作纯粹的技术命题去攻关，探索数据形态、模型交互，以及如何将廉价数据高效转移到模型学习中。Sarah 在初次面谈时被这种以结果为导向的极客思路彻底打坚，当场决定投资。 从斯坦福地下室里用纸板拼装的模型起步，到如今能够自主研发制造软硬件垂直整合的机器人，整个团队在不到两年的时间里经历了上百次硬件迭代与真实世界任务测试。业内原本普遍认为“家庭通用半人型机器人”遥不可及，但该团队预计在今年年底就会推出面向家庭的 Beta 测试版，最迟也不过明年。机器人在真实环境中的演进速度令人瞠目结舌，让整个行业彻底从怀疑“能不能实现”，转向了确信“只是时间早晚问题”。 ```mermaid flowchart TD accTitle: 具身智能数据突破路径 accDescr: 展示从传统机器人数据困境到低成本采集与模型泛化的技术路径 subgraph Problem["传统困境"] A["缺少机器人物理交互数据"] --> B["获取数据成本极其高昂"] end subgraph Innovation["核心技术突破"] B --> C["设计低成本、高效率的采集装置"] C --> D["覆盖真实世界的多样化任务分布"] end subgraph Outcome["落地与迭代"] D --> E["模型高效迁移与快速训练"] E --> F["数百次软硬件协同迭代"] F --> G["进军家庭场景的半具身/通用机器人"] end ``` --- ### `22:15` 投资决策机制：从直觉出发，以备忘录校准 当主持人问及她在初次会面就拍板投资是否属于常态时，Sarah 剖析了 Conviction 内部真实的投资决策机制。她坦言自己在看人方面高度依赖直觉：一旦了解创始人的技术积累，并在面对面交流中感受到其清晰的构想，她往往能在几分钟内形成判断。内部有一个 1 到 10 分的评分系统，在许多心动的案子上，她第一面就会打出 8 分甚至 9 分。 但从最初的直觉到最终签署条款，中间的过程是一场严密的反向自我推翻。Sarah 会花接下来的数天甚至数周时间，极力寻找自己认知中的盲区：“在哪些地方我的判断可能是错的或不完整的？关于这个技术或赛道，其他专家究竟相信什么？他们真的是我认为的那种顶尖人才吗？”早期风险投资涵盖生物、国防、具身智能、法律等极广的跨界领域，投资人不可能样样精通，因此必须快速扎根现实建立认知基准。 在合伙人协作上，Sarah 沿用了在 Greylock 养成的高标准备忘录（Memo）文化。哪怕早期基金只有她一个人，她也会把完整的投资备忘录写好，发给像 John Lilly 或 Dylan Field 这样信任的朋友进行逻辑挑战。她强调：“虽然我自己能独立做决策，但我深信他人能让我变得更强，我需要别人挑刺来倒逼自己反思。”如今团队中的合伙人（如 Mike、Bella、Pranav）有着互补的决策方式：有些人是从零开始慢慢推理、逐步建立信心，而 Sarah 则是先形成强烈直觉再反推验证漏洞，团队共同把关“究竟是什么事实才能改变当前的判断”。 - **直觉与逻辑交织**：Sarah 并不赞同盲目迷信创始人声誉。即使是像 Bret Taylor 这样顶级声誉的创业者，她之所以毫无保留地支持，是因为看到了其长久以来无懈可击的产业逻辑（Industrial Logic）与反复被验证的正确判断；如果缺乏对业务本身的深层认知，就无法对创始人的判断力做出真正准确的评估。 - **打分体系与否定测试**：初面给出高分后，后续的投研流程聚焦于“寻找反例”和“破除信息盲区”，而非寻找证据去自我证实。 --- ### `26:13` 时间分配的艺术：从疯狂扫会到高密度深度学习 谈及目前作为基金掌门人的日常日程与精力切片，Sarah 分享了自己清晰的时间分配比例： - **60% 的时间投入投后与赋能**：这是最大的一块占比，包括协助被投公司核心人才招聘、高层战略推演、链接外部生态资源以及协助后续融资。 - **约 20% 的时间看新项目**：每周只深度见 4 到 6 家新公司。这与她早期在大型机构时头几个月猛看 500 家公司形成了鲜明对比；经过长期的校准，她已经能够用极高的信噪比敏锐识别出真正的顶尖项目，无需再靠海量会议刷存在感。 - **约 20% 的时间用于学习与拓展认知边界**：包括帮助合伙人复盘、与前沿研究员长谈，甚至跨界与二级市场（公开市场）投资人、大型制药公司（Pharma）高管沟通，了解传统产业对 AI 转型的真实预期。她特意在日程表里保留充足的空白，专门用来满足好奇心、吸纳不同资产类别的思考逻辑。 --- ### `28:15` 募集首期基金：对 LP 讲实话的下注 在筹备 Conviction 第一期基金时，曾有一位私募股权（PE）领域的投资人朋友建议 Sarah：“你必须给 LP 讲一个高度包装、差异化的故事，把漏斗模型中每一步具体怎么做严丝合缝地画出来。”然而 Sarah 却选择了截然相反的坦诚策略。她深知，在技术剧烈动荡的前沿市场，在未进入实际作战前就声称自己全盘知晓是虚伪的：“老实讲，我还没完全弄清楚全部细节，正因如此我才需要募集资金，去市场上通过真实的实践与碰撞把路走通。” 在发给潜在出资人（LP）的材料中，她只准备了一份简洁的两页纸（Two-pager），列明自己的投资历史、战绩，以及最核心的底层主张：她极其擅长识别出卓越非凡的非凡人才，并能成为这些人才最坚定的盟友和最有力的招聘帮手。只要把这种顶尖敏锐的投资判断力、号召力与拼命执行的决心结合在一起，就构成了最具爆发力的起点。 这种拒绝迎合、不做表面文章的风格虽然吓退了部分偏好规整叙事的机构，但也赢得了那些真正懂行的初期 LP。那些在早期下注 Conviction 的出资人，在向其投委会汇报时写下的核心逻辑往往直接而纯粹：“她会拼尽全力去执行到底，我们拭目以待。” ### `30:00` 投资机构的创业表达与独立思考 在投资理念的沟通上，Sarah Guo 认为开诚布公能让事情变得纯粹得多：“如果你告诉人们你打算做什么、你内心真正相信什么，生活会简单得多。”作为一名投资人，每当有人能用严密的逻辑说服她“这个世界的运转方式其实与你原先所想的不同”时，她会产生一种难以言喻的兴奋感。过去四年里，她接触了大量极具创意的投资经理。与传统风险投资人（VC）不同，她尤为推崇那些带有强烈企业家精神的投资管理者。 Sarah 坦言，自己并不想完全复制这些同行所建立的机构形态，但她由衷钦佩一些先锋管理者的创新精神。例如 Coatue 的 Philippe Laffont 和 Thomas Laffont 兄弟，以及 Thrive Capital 的 Josh Kushner。他们对待投资管理业务的方式极具创造力，并且不断鼓励同行打破旧有范式：投资机构本身就是一种表达个人观点的产品形态，优秀的投资人完全应该通过自己的基金载体去践行崭新的理念。 ### `30:49` 父母的言传身教与抵制创投圈的零和博弈 谈及成长背景，Sarah 的父母都是创业者。尽管双亲始终在极其拼命地工作，但在她的童年和少年时期，她从未感觉父母缺席过自己的成长。她从小就非常独立，这种性格帮助她很早便意识到：父母虽然能让你感受到你是他们世界的中心，但他们同时也是拥有自我追求、渴望在其他领域实现价值的完整人类。在家庭价值观中，诚信、团队凝聚力与“独立思考”被置于最高位置，而父母甚至将独立思考上升到了道德准则的高度——“你绝不能在乎别人怎么看你”（"You cannot ever worry about what other people think"）。 虽然 Sarah 自认在绝大多数时候都能坚守特立独行，但她坦承自己有时仍会在意激起创投圈同行的竞争戒心。她骨子里是一个温和、推崇正和博弈的人，而传统 Series A、Series B 风投机构常常充斥着残酷的零和博弈： > “那种‘我必须拥有这家公司 18% 到 25% 的股权并拿下一个董事会席位，而你一分钱都别想分到’的姿态，根本不利于行业合作。” 相比之下，她更向往公开市场那种共享认知、正和共赢的氛围——投资人乐于分享自己发现的最佳商业机会，甚至号召其他人一起跟进。比如她非常渴望公开探讨为什么游戏与娱乐行业将被 AI 彻底颠覆，而当下的市场对此依然极度低估。 对 Sarah 而言，最深层的灵感始终来源于早年目睹父母创业的经历：“那是‘我们对抗巨头’（It's us against the man）的热血过程。大公司体量庞大，试图掐死初创企业，但只要技术更优、产品更强、能够切实满足客户需求，小团队依然能够胜出。”她对敢于从无到有创造未来的创业者抱有近乎本能的敬意。正因如此，她对当下后 Z 世代创投圈中滋生的犬儒心态深恶痛绝：部分年轻人错误地以为创业全靠营销包装、搞关系（nepotism）和在社交媒体（Twitter/X）上博眼球。Sarah 强调，人脉和品牌固然真实存在，但唯有死磕产品价值、善待客户与伙伴，才能在商业世界中经受住真正的考验。 ### `34:38` 开源 AI 的必然性与生态安全考量 围绕非美国本土前沿开源模型的崛起，业内正展开激烈论辩。面对中国团队的高水平开源成果，以及欧洲的 Mistral、Poolside 等模型，以及像 BaseTen 这样承载海量开源模型推理的基础设施平台，Sarah 明确指出：现实是“猫已经从袋子里跑出来了”（"The cat is out of the bag"），全球范围内的开源模型已经全面铺开且性能逼近前沿，这一技术扩散的客观趋势无法逆转。 ```mermaid flowchart TD accTitle: 开源 AI 扩散与经济落地架构 accDescr: 展示前沿闭源 API 的落地瓶颈与开源模型如何通过民主化技术支撑庞大实体经济 subgraph FrontierConstraints ["前沿闭源 API 的现实瓶颈"] A["高昂调用成本 (Cost)"] B["数据隐私与敏感性 (Privacy)"] C["响应延迟与容量瓶颈 (Latency)"] end subgraph OpenSourceEcosystem ["开源模型与实体扩散"] D["全球前沿开源模型<br/>(美国、中国、欧洲 Mistral/Poolside 等)"] E["后训练与 Harness 工具民主化"] F["企业自控数据与算力命运"] end subgraph RealEconomy ["千行百业的落地场景"] G["复杂业务分支与长尾需求"] H["国防与网络安全防御"] I["生物学与科学研究"] end FrontierConstraints -->|"推动企业转向"| D D --> E E --> F F --> G F --> H F --> I ``` 从经济规律来看，如果各行各业仅仅依赖头部前沿实验室的专有闭源 API，很快就会撞上成本过高、延迟过大以及数据隐私敏感的墙。随着 AI 应用的深化，开源模型是实现技术平权与广泛商业渗透的基石。对于所谓的国家安全担忧，Sarah 提出了务实且犀利的见解： - **禁止开源只会自缚手脚**：如果在本土强行压制或禁止开源模型，受到惩罚和限制的只会是遵纪守法的本国正规企业，拖慢本土创新；而真正的网络攻击者和对抗性对手根本不会遵守禁令。 - **双重用途的本质**：前沿模型既能用于网络防御和生物医药研发，同样具备被攻击者滥用的潜力。与其凭空猜测模型是否存在所谓的“后门”，不如构建极其严格的前沿安全测试矩阵，用实证研究替代空洞的恐慌炒作。 - **实验室无法穷尽真实世界**：正如 Sam Altman 所描绘的“便宜到无需计量的智能”（intelligence too cheap to meter），这一未来必须依靠开源生态来承载。前沿实验室的研究员无论多么聪明，都无法凭空想象出庞大实体经济中千奇百怪的现实业务场景；只有让模型工具链和后训练技术普惠化，由无数企业在廉价基础设施上自由构建，AI 的价值才能真正释放。 ### `39:01` 廉价智能并非历史必然与美国制造业的现实挑战 针对主持人提出的历史警示——美国曾预言核能将带来“便宜到无需计量的电力”，但现实中因监管僵化与建设成本失控（如西屋 AP1000 反应堆项目的坎坷），这一宏愿在本土化为泡影，反观中国却在核电建设上高歌猛进——AI 是否也存在“丰沛廉价智能并未如期而至”的可能？ Sarah 毫不犹豫地给出了肯定的回答：她完全能够想象出一个美国在智能基建竞争中掉队、未能获得廉价丰沛算力的严峻未来。这不仅是一场技术竞争，更是关乎国家命脉与工业安全的生死战。 Sarah 强调，美国要想重建本土工业基础、打造具备弹性的供应链，面临着劳动力极其昂贵且制造业技能严重断代的残酷现实；在不大幅引进外来工人的前提下，唯一的生路就是依靠高强度的自动化与智能化。如果廉价且强大的 AI 无法如期落地，重振制造业的构想就只是空中楼阁：“如果不用自动化，那么到底由谁来生产这些关键物资？”（"Who's going to produce this stuff?"）在她看来，确保智能基础设施的低成本与充足供应，是保障现代工业体系与国家竞争力不可妥协的底线。 ### `40:00` 算力独立性与能源基建的现实瓶颈 美国人不应该、也不希望拿着每小时 13 美元的薪资去从事缺乏尊严与人性的机械劳动，但这并不意味着美国在先进产业中的竞争力是天然存在的。全球竞争力的维系需要主动做出战略抉择。当前一种极具破坏力的现实风险在于，公众因理性担忧 AI 对就业的冲击，或反感财富与租金被少数几家超级科技巨头攫取，从而抗拒沦为“永久底层阶级”。这种情绪若与反资本主义思潮合流，极有可能直接拖慢能源、电网、数据中心以及工业制造产能等关键基础设施的建设步伐。 而在所有这些要素中，最核心的硬约束正是算力。一旦失去算力储备，无论在经济上还是科技上都将彻底丧失竞争力，更谈不上独立自主。“算力独立性”（Compute Independence）正在成为国家安全和产业界的核心命题。从一个摆满 GPU、配备高密度散热与庞大供电系统、负责支撑训练与推理任务的现代数据中心往上游倒推，整个链条与能源独立的逻辑如出一辙：它高度依赖脆弱且复杂的全球供应链。许多最核心的技术组件就像是一道极其狭窄的筛网，完全受制于地缘局势敏感、对美国及盟友而言并不绝对安全的区域。例如，半导体制造中某种不可或缺的特种玻璃基板，几乎被极少数供应商甚至单一巨头牢牢垄断。 ```mermaid flowchart TD accTitle: 算力独立性供应链倒推图 accDescr: 从最下游的应用端倒推至最上游的原材料与设备瓶颈。 subgraph AppDownstream ["应用与计算中心"] Apps["AI 终端应用 / 训练与推理"] DataCenter["数据中心基础设施<br/>(高密电力 / 冷却系统)"] Apps --> DataCenter end subgraph HardwareLayer ["算力硬件与能源"] Chips["先进制程芯片 / 替代架构"] Energy["核能 / 清洁能源供应"] DataCenter --> Chips DataCenter --> Energy end subgraph Chokepoint ["上游关键瓶颈"] Fab["先进制程晶圆代工<br/>(台积电等先进制程产能)"] RawMat["特殊工业原料与设备<br/>(高规格玻璃基板 / 关键耗材)"] Chips --> Fab Fab --> RawMat end ``` 面对这一现状，类似 Jacob Helberg 倡导的“Pax Silexica”战略正在兴起——审视供应链的每一个环节，通过政策引导和资本投入寻找具备冗余度的独立路径。这并不意味着所有环节都必须搬回美国本土，比较优势定律依然成立，但建立多元化的供应来源是所有参与者的当务之急。在这一领域，Conviction 已经围绕多个节点展开布局：从填补数据中心与机器人领域的重度蓝领劳动力缺口，到押注核能发电项目，再到投资新型替代芯片架构。至于纯粹的数据中心施工、光伏与电池安装承包等重资产业务，Sarah 坦言其本质介于重度运营、工程地产与复杂的项目融资之间，尽管团队进行了非常深入的尽调，但目前尚未出手，因为作为风险技术投资人，她依然在寻找具备长期防御壁垒的技术型产品资产。 - 算力独立性的核心并非关门自造，而是消灭单点故障，构建供应链冗余度与多源供应格局。 - 芯片制造上游存在极窄的技术瓶颈（如关键的特殊玻璃原材料），供应链脆弱性堪比能源危机。 - 资本在重资产数据中心建设与硬科技底层资产之间存在分野，科技风投更看重具备持续复利的技术产品本身。 --- ### `43:14` 改变认知的内部辩论：半导体与非共识市场 在 Conviction 内部，最激烈的讨论往往围绕“传统认知何时失效”展开。正如 Sarah 与 Elad Gil 共同主持的播客《No Priors》所探讨的核心逻辑，许多投资人过去深信不疑的市场铁律，在新的技术范式下正在被彻底推翻。半导体投资就是一个典型案例：在过去数十年的风投历史上，除了少数偶发奇迹，投资芯片初创公司几乎是一门“烂到极点”的赔钱生意，几乎所有老牌投资人都会告诫年轻后辈不要涉足。 然而，团队中的年轻合伙人开始重新审视这一赛道，并敏锐地指出市场逻辑已发生根本性逆转。今天，全球对专用算力加速芯片的需求出现了前所未有的大规模集中爆发，连最顶级的买家自身也迫切渴望摆脱对单一晶圆产线的绝对依赖。“大家不可能永远挤在同一条晶圆代工产线上排队”，这种前所未有的供应链独立诉求从根本上改写了初创芯片公司的风险回报公式。 Conviction 团队在大的技术演进方向和值得解决的宏大问题上往往高度一致，真正的交锋在于：目标市场是否能够支撑起典型的风险投资回报分布（Power-law Distribution）？风投模式在历史上高度偏爱纯软件，而航天、太阳能、储能电池、核裂变、涡轮机制造、机器人以及生物科技等硬核领域，从来都不是十年前风投机构舒适区内的典型软件市场。评估这些市场能否孕育出风险投资级别的百倍回报，构成了公司内部最活跃的智识思辨。 - 传统认知认为半导体投资回报极低，但在超级买家强烈需求与供应链脱钩焦虑的双重推动下，其风险收益模型已发生质变。 - 早期风投的辩论重点不再是“问题是否真实存在”，而是“该商业环境是否允许初创企业捕获超额价值”。 --- ### `45:27` 生物科技的质变：从卖药逻辑到模型驱动 在生物计算（Computational Biology）领域，实打实的商业数据彻底颠覆了 Sarah 过去的怀疑态度。在长达五六年的时间里，生物医药界存在一个坚不可摧的行业共识：服务制药行业赚不到大钱，纯做软件或平台型公司根本走不通；生物科技初创公司想要赚钱，唯一的路径就是自己研发药物分子，然后通过授权交易（License/Biobucks）分阶段兑现价值，并决定在漫长且高死亡率的临床管线中承担多大的风险。 这种传统模式塑造了独特的生物医药风投形态：顶级机构通常寻找高校明星实验室的首席科学家（PI），直接拿走 40% 的股权并强力介入公司组装，将科研成果快速包装成药物候选分子，即便明知绝大部分项目最终都会在临床试验中折戟沉沙。在传统生物风投眼中，软件投资人试图向药企卖软件工具的行为近乎天真。但大模型的爆发正在打破这一行业铁律。作为 Chai Discovery 的首家机构投资人（First Check），Conviction 见证了这家公司与多家全球排名前十的跨国制药巨头签下极其重磅的商业合作，直接切入核心研发管线。 这种认知的转变并非来自于某种先验的顿悟，而是在未知中勇敢下注后迎来的客观现实。当一家成立不久的公司能够凭借底层模型直接拿下千万美元级别的商业大单，并且当直接与客户企业内部的一线科学家和研发部门主管深度沟通时，客户能够清晰明确地指出该工具为实验研发带来的巨大增益与商业价值。虽然物理世界的临床实验验证、药物安全性筛查以及极其严格的监管审批依然是无法缩短的物理周期，但整个制药行业拥抱 AI 的速度已经远超外界预期。未来只要出现第一款完全由 AI 深刻改变研发轨迹并成功获批的新药或新适应症，整个生物医药领域的投资浪潮就将被彻底引爆。 - 传统生物医药投资信奉“唯有做药分子才有价值”，向药企销售软件被长期视为无法规模化的非主流路径。 - Chai Discovery 凭借基础生物模型拿下大药企千万美元级合同，证明了纯平台与工具模型在研发端的真实变现能力。 - 物理实验与监管安全虽然是不可逾越的客观规律，但 AI 对前期研发管线的指数级加速已得到一线科学家的全面认可。 --- ### `48:58` 何谓“Conviction”：极端信念与个人担责 将基金命名为“Conviction”（信念），本身就带有一种强烈的追求与愿景。在早期风险投资最纯粹、最经典的形态中，投资人在企业微末之时以极其早期的姿态入局，建立举足轻重的股权头寸，此后始终坚定持有，陪伴创始人一路前行，直至公司最终取得巨大成功、被收购或是走向终局。这种机制蕴含着极具美感的利益绑定与纯粹性。 科技史上最伟大的一批企业——无论是 Figma、Notion 还是 Rippling——在其创立最初的数年里，商业前景往往都极为模糊，甚至充满了不确定性。没有任何一个优秀的投资人会在下注时指望一家公司要熬上四五年才能找到产品市场契合点（PMF），但伟大的早期投资恰恰在于：当市场因创始人的背景或非主流的方向而普遍充满疑虑时，你能够建立起高度独立的非共识判断，主动搁置怀疑，以百分之百的信念全力支持对方，直到设想被证实或证伪。 ```mermaid flowchart LR accTitle: 投资决策与信念演进逻辑 accDescr: 展示从非共识洞察到承担个人决策责任，最终形成陪伴企业成长的信念机制。 A["非共识洞见<br/>(市场未被定价的盲区)"] --> B["搁置怀疑 / 承担下注风险<br/>(Suspend Doubt)"] B --> C["合伙人个人全权负责<br/>(Individual Ownership)"] C --> D["长期深度陪伴与支持<br/>(Conviction 直至验证成功)"] ``` Sarah 极其推崇“投资人个人对决策承担完全责任”（Individual Ownership）。她完全无法认同某些风投机构所推行的“集体负责制”或“非个人所有制”投资文化： > “我跟一家投资机构的负责人聊天，他说他们机构的投资项目没有个人归属。这对我来说简直是天方夜谭！你怎么能用这种方式经营投资业务？必须有人站出来为此负责。” 真正的合伙人机制必须是一场团队运动，但每一次挥棒必须由具体的合伙人亲自决断：“你真的相信这个项目吗？说服我们，我们又该如何协助你做出最终决定？”在风险承担上，Sarah 坦言自己并不像 Founders Fund 的朋友们那样天生具有强烈的反叛逆向（Contrarian）本能；她所坚持的本质，是敢于建立自己独立、真实的深思熟虑，并拥有完全不在意他人世俗评价与外界杂音的心理韧性。 - 早期投资的真谛在于面对模糊性时主动搁置怀疑，以坚定的信念陪伴公司穿越长期的验证周期。 - Conviction 严苛推行合伙人个人负责制，坚决反对稀释决策责任的集体委员会决策文化。 - 真正的风险承担不是刻意为了逆向而逆向，而是忠于自己推导出的事实真相，彻底摆脱从众心理的束缚。 ### `50:00` 寻找真实、过滤噪音与非对称下注 在面对市场中甚嚣尘上的主流叙事，或行业关键巨头各自发表的宏大宣言时，投资的核心逻辑其实非常简单：探寻“真实”（The Truth）。将这一逻辑与风险承受力联系起来看，投资的本质就是在发现被市场错误定价的真实后，坚定地持有它。投资人真正追求的是“非对称信息”（Asymmetric Information），以及在所有人尚未想明白之前坚守自己观点的定力。为了构筑这种深度的认知护城河，关键在于将精力分配给那些真正让事情发生的人，而不是陷入宏观理论的无休止争论。 Sarah Guo 将最有价值的交流对象划分为三类：正在做出突破、带来颠覆性惊喜的创业者；在未知前沿探索的顶级研究员；以及非常聪明、持有与她截然相反信念的人。她坦承自己曾犯过直觉上的重大失误——例如 Suno 的联合创始人兼 CEO Mikey Shulman 正在打造一款极具潜力的音乐生成产品。当时一位共同的投资人朋友引荐她投资，她却愚蠢地拒绝了，理由是“我不认为会有那么多人想做音乐，虽然我自己会做，但这能变成社交网络吗？会有多少消费需求？”然而，事实证明她的直觉完全错了，人们对 AI 工具在表达、娱乐和创作上的渴望远超预期。与这类正在创造不可思议行为模式的创始人交流，所学到的认知远胜于空谈。 ```mermaid flowchart TD accTitle: Conviction Information Filtering Framework accDescr: Diagram illustrating how Conviction filters out circular macro noise to focus on frontline diffusion. subgraph MacroNoise ["行业噪音 (避免过度消耗精力)"] N1["大模型实验室策略猜想"] N2["宏大框架：'哪一层会赢'"] N3["应用层能否存活的循环论证"] end subgraph FrontierTruth ["真实前沿 (重点精力投入)"] T1["一线创业者创造的新用户行为 (如 Suno)"] T2["前沿研究探索的突破性方向"] T3["企业内部打造的 AI 自主杠杆"] end subgraph InvestmentEdge ["投资非对称优势"] E1["识别被错误定价的真实"] E2["在众人理解之前坚定持有"] E3["聚焦 AI 扩散的剩余 99%"] end FrontierTruth --> InvestmentEdge ``` 相比之下，投资圈往往陷入一种循环论证：天天揣摩大模型实验室的下一步战略是什么、应用层企业到底能不能活下来。在 Sarah 看来，这类讨论对实际决策几乎没有任何指导意义。大模型巨头自然会选择在安全、高利润的路径上逐步扩张（如从 ChatGPT 到广告、再到编程辅助以及未来的 Co-work 协同工具）。投资人需要具体评估这些尝试的实际竞争力与边界，而不是花费大量精力去推演一个“究竟哪一层架构能通吃一切”的宏大战略框架。如果当前 AI 的普及还处于 1% 的极早期阶段，那么探究未来 99% 的扩散（Diffusion）究竟如何发生，才是最值得下注的方向。 --- ### `54:16` 焦耳悖论与未来一年的技术演进 展望未来一年的技术演化，在快节奏的“反向狗年”发展速度下，Sarah 期待看到“杰文斯悖论”（Jevons Paradox）在现实世界中的全面显现。正如同软件工程领域所发生的范式转移一样，当越来越多的智能体（Agents）和自动化产品开始更高效地处理各类繁琐平庸的日常事务时，整体的产出与创造力并不会减少，反而会催生出巨大的增量需求。各个领域的公司都在以惊人的速度推进自身迭代。 这种现象已经在 Conviction 投资组合的企业内部清晰可见。例如，某家服务众多客户的被投企业，其原本的传统市场营销工作（如销售物料制作、内容生成）需要专人负责，而该公司的营销负责人极具通过工具撬动自身杠杆的意识，索性直接为公司构建了一整套“自主化运作的虚拟市场部”。在所有职能部门中，随着学习速度的提高和平庸工作的减少，节省下来的时间都会被重新配置到更高价值的任务中。正如英伟达 CEO 黄仁勋（Jensen Huang）的核心智慧所揭示的那样：AI 的普及不会让人无事可做，反而会让大家更加忙碌并创造更多就业，关键在于必须为每个人提供使用这些前沿工具的途径与教育培训。 - **杰文斯悖论的重现**：平庸日常工作的边际成本大幅下降后，企业和个人并没有减少工作量，而是借助更高的生产力追求前所未有的业务广度与推进速度。 - **自主化部门的出现**：非工程师背景的职能主管（如市场负责人）通过自建 AI 工作流，一个人便能实现全功能市场部的产出效率。 - **生产力驱动的劳动再投入**：面对“使用 AI 更高产后是否会更少工作”的提问，Sarah 的回答是“我会工作得更多”。 --- ### `56:46` 硅谷最纯粹的善意：在毫无凭证时选择相信 谈及人生中他人所给予的最珍贵的善意，Sarah 动情地回顾了硅谷独有的包容文化：这里汇聚着众多成就卓越的人，但他们绝不迷信所谓的光鲜履历与出身。只要一次深入的交谈让他们感受到你能提供帮助、拥有独到见解，或者单纯觉得你很有潜力，他们就愿意毫无保留地在一份关系和一个人身上押注。这种基于“人本身和想法本身的纯粹评估”来建立信任的机制，在其他任何行业生态中都极其罕见。 回顾早期的职业生涯，Greylock 的前辈们——Asheem Chandna、Aneel Bhusri、Joseph Ansanelli 以及 Reid Hoffman 在 Sarah 仅有 23 岁时就毅然雇用了她。年轻投资人往往容易沦为行业嘲讽的靶子（“这帮小毛孩能懂什么？为什么让年轻人掌管资金？”）。然而，Sarah 坚信通过第一性原理思考和向优秀者模仿，年轻人在任何具体事务上都能迅速追求卓越。正是这些导师倾注时间、不计回报地教导她如何成为一名真正的投资人，给了她最宝贵的起步信任。 而在她决定离开成熟平台、创立 Conviction 时，包括 Ravi Gupta（现任 Ithaca 联合负责人）、Dylan Field、Elena Donio、John Lilly 等朋友，以及首批出资的 LP 们，都毫不犹豫地给予她最坚定的支持与鼓励。尽管即便没有这些鼓励，她也依然会坚定地踏上创业之路，但这种“在毫无现实证据支撑前便给予信任”（Faith / Belief without evidence yet）的信念，是推动世界运转的最美好力量。 - **年轻视角的价值**：23 岁入行早期风险投资，从专注为他人创造成功起步，通过第一性原理迅速积累专业能力。 - **摒弃资历偏见的文化**：硅谷顶尖创业者与投资人愿意纯粹根据“想法的质量”与“个人的潜质”进行投入，使没有特殊背景的新人同样拥有崭露头角的机会。 - **Conviction（信念）的真谛**：投资与创业的底色，归根结底是对人本身的信仰；在证据尚不齐备的混沌中托付信任，成就了真正伟大的事业。 ## Notable quotes > `00:00` 我总跟自己说必须竭尽全力踩死刹车，可现实中我却开着车以 90 英里的时速在狂飙，根本停不下来。 — Sarah Guo 引述投资人朋友 > `03:31` 我非常明确地不希望看到那样的未来，而且我认为我们绝不会落入那样的境地。 — Sarah Guo > `10:00` 我自己觉得距离 AGI 还有两年这句话已经说了十年 — Andre Karpathy > `22:15` 虽然我自己能独立做决策，但我深信他人能让我变得更强，我需要别人挑刺来倒逼自己反思。 — Sarah Guo > `28:15` 老实讲，我还没完全弄清楚全部细节，正因如此我才需要募集资金，去市场上通过真实的实践与碰撞把路走通。 — Sarah Guo > `30:00` 如果你告诉人们你打算做什么、你内心真正相信什么，生活会简单得多。 — Sarah Guo > `30:49` You cannot ever worry about what other people think — Sarah Guo 的父母 > `30:49` 那种‘我必须拥有这家公司 18% 到 25% 的股权并拿下一个董事会席位，而你一分钱都别想分到’的姿态，根本不利于行业合作。 — Sarah Guo > `34:38` The cat is out of the bag — Sarah Guo > `39:01` Who's going to produce this stuff? — Sarah Guo > `48:58` 我跟一家投资机构的负责人聊天，他说他们机构的投资项目没有个人归属。这对我来说简直是天方夜谭！你怎么能用这种方式经营投资业务？必须有人站出来为此负责。 — Sarah Guo > `50:00` 我不认为会有那么多人想做音乐，虽然我自己会做，但这能变成社交网络吗？会有多少消费需求？ — Sarah Guo > `54:16` 我会工作得更多 — Sarah Guo ## Mentioned Sarah Guo, Patrick O'Shaughnessy, Conviction, OpenAI, Anthropic, Harvey, Winston Weinberg, Gabriel Pereyra, Activision Blizzard, Andre Karpathy, Sunday Robotics, Tony Zhao, Cheng Chi, Zipeng Fu, Physical Intelligence, Greylock, John Lilly, Dylan Field, Bret Taylor, Coatue, Philippe Laffont, Thomas Laffont, Thrive Capital, Josh Kushner, Mistral, Poolside, BaseTen, Sam Altman, Jacob Helberg, No Priors, Elad Gil, Founders Fund, Chai Discovery, Figma, Notion, Rippling, Suno, Mikey Shulman, Jensen Huang, Asheem Chandna, Aneel Bhusri, Joseph Ansanelli, Reid Hoffman, Ravi Gupta, Elena Donio ## Video description My guest today is Sarah Guo, founder and managing partner of Conviction, the venture firm she built to back AI-native companies from their earliest days.&nbsp; Sarah has become one of the most sought-after early-stage investors in AI, often the first check into the companies defining the frontier.&nbsp; In this conversation, we go inside that frontier: what the small group of people actually building AI believe right now, why some of the field's best researchers are wrestling with their own sense of purpose, and how close we are to robots in the home and a genuine acceleration in scientific discovery.&nbsp; At the center is Sarah's conviction that no single company will own the future of AI, and what that means for founders, investors, and anyone allocating their time and resources in a world moving this fast. Our managing editor Dom Cooke wrote a profile of Sarah for Colossus, "Sarah's Wager," on how she built the firm closest to the AI frontier and why she's now betting against its biggest companies.&nbsp; Please enjoy this conversation with Sarah Guo. For the full show notes, transcript, and links to mentioned content, check out the episode page ⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠here⁠⁠⁠⁠⁠.&nbsp; ----- Become a Colossus member to get our quarterly print magazine and private audio experience, including exclusive profiles and early access to select episodes. Subscribe at ⁠colossus.com/subscribe⁠. ----- ⁠Ramp’s⁠ mission is to help companies manage their spend in a way that reduces expenses and frees up time for teams to work on more valuable projects. Go to⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠ ⁠ramp.com/invest⁠⁠ to sign up for free and get a $250 welcome bonus. ----- Trusted by thousands of businesses, ⁠Vanta⁠ continuously monitors your security posture and streamlines audits so you can win enterprise deals and build customer trust without the traditional overhead. Invest Like the Best listeners get a special offer of $1,000 off Vanta when you go to ⁠vanta.com/invest⁠.&nbsp; ----- WorkOS⁠ is the infrastructure B2B and AI-native companies use to sell to enterprise. It covers everything enterprise security requires: SSO, SCIM, RBAC, Audit Logs, AI governance, and more. Trusted by 2,000+ fast-growing companies, including OpenAI, Anthropic, Cursor, and Vercel. ----- Rogo is the AI platform for finance. They're building agents for Wall Street that are trained to understand how bankers and investors actually do work: from diligence and modeling, to turning analysis into deliverables. To learn more, visit rogo.ai/invest. ----- ⁠Ridgeline⁠ has built a complete, real-time, modern operating system for investment managers. It handles trading, portfolio management, compliance, customer reporting, and much more through an all-in-one real-time cloud platform. Visit⁠ ⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠ridgeline.ai⁠. ----- Editing and post-production work for this episode was provided by The Podcast Consultant. Timestamps: (00:00:00) Welcome to Invest Like the Best (00:02:16) Investing Without a Backtest (00:03:31) The AI Wager (00:06:16) Building the Best Investment Firm (00:08:37) Finding Non-Obvious AI Opportunities (00:11:00) The Frontier AI Talent Race (00:13:50) Compute as the Constraint (00:19:10) The Future of Robotics (00:22:15) Making Investment Decisions (00:26:13) How Sarah Spends Her Time (00:28:15) Raising a Venture Fund (00:30:49) Lessons From Her Parents (00:34:38) The Case for Open Source AI (00:39:01) Abundant Intelligence Isn't Inevitable (00:40:55) Compute Independence (00:43:14) Debates Inside Conviction (00:45:27) AI's Opportunity in Biology (00:48:58) Why Conviction (00:50:31) Finding Truth and Taking Risk (00:54:16) What Changes in the Next Year (00:56:46) The Kindest Thing [Source](https://traffic.megaphone.fm/CLS7890997457.mp3)

[0:00] Sarah Guo 引述投资人朋友: 我总跟自己说必须竭尽全力踩死刹车，可现实中我却开着车以 90 英里的时速在狂飙，根本停不下来。 [3:31] Sarah Guo: 我非常明确地不希望看到那样的未来，而且我认为我们绝不会落入那样的境地。 [10:00] Andre Karpathy: 我自己觉得距离 AGI 还有两年这句话已经说了十年 [22:15] Sarah Guo: 虽然我自己能独立做决策，但我深信他人能让我变得更强，我需要别人挑刺来倒逼自己反思。 [28:15] Sarah Guo: 老实讲，我还没完全弄清楚全部细节，正因如此我才需要募集资金，去市场上通过真实的实践与碰撞把路走通。 [30:00] Sarah Guo: 如果你告诉人们你打算做什么、你内心真正相信什么，生活会简单得多。 [30:49] Sarah Guo 的父母: You cannot ever worry about what other people think [30:49] Sarah Guo: 那种‘我必须拥有这家公司 18% 到 25% 的股权并拿下一个董事会席位，而你一分钱都别想分到’的姿态，根本不利于行业合作。 [34:38] Sarah Guo: The cat is out of the bag [39:01] Sarah Guo: Who's going to produce this stuff? [48:58] Sarah Guo: 我跟一家投资机构的负责人聊天，他说他们机构的投资项目没有个人归属。这对我来说简直是天方夜谭！你怎么能用这种方式经营投资业务？必须有人站出来为此负责。 [50:00] Sarah Guo: 我不认为会有那么多人想做音乐，虽然我自己会做，但这能变成社交网络吗？会有多少消费需求？ [54:16] Sarah Guo: 我会工作得更多

[3:31] Sarah Guo: I believe in like the great man and great woman theories of history. If you have very high agency people in all of these places and they have the correct risk capital... yeah, you change outcomes. [3:31] Sarah Guo: The extreme point of view that some folks may have... of the owner of one, two, three frontier models consumes the economy. I do not want that future very clearly. [11:00] Sarah Guo: Andre Karpathy will actually, in a very self-aware way, say, 'I thought it was two years away for about 10 years.' And he thinks it again, to be fair. [28:15] Sarah Guo: If you tell people what you are going to do, life is much simpler, and when you actually state what you believe, life is much simpler.

[0:00] Ramp is the only platform built to make your finance team leaner, faster, and better, saving businesses 5% annually on average so you can stay focused on growth. Ramp customers grow revenue 3.2 times faster than the average American business. Visa, Vercel, Cursor, Stripe, Notion, 11Lab, Shopify, and 70,000 other businesses all now run on Ramp. Mine does too, and so should yours. Learn more at ramp.com slash invest. ready work using your firm's own templates, context, and standards. Send Felix an email like, take these comments and turn them for me. Or update my tracker with the context of these emails. And Felix sends back finished PowerPoint decks, Excel models, and sourced research. Felix works the way your team already does, delivering work quickly and accurately around the clock. Learn more at rogo.ai slash Felix. OpenAI, Cursor, Anthropic, Perplexity, and Vercel all have scale, you have to deliver on core capabilities like SSO, SCIM, RBAC, and audit logs. Instead of spending months building these mission-critical capabilities yourself, you can just use WorkOS APIs to gain all of them on day zero. That's why so many of the top AI teams you hear about already run on WorkOS. WorkOS is the fastest way to become enterprise-ready and stay focused on what matters most, your product. Visit workos.com to get started. Hello and welcome, everyone. I'm Patrick O'Shaughnessy, and this is Invest Like the Best. This show is an open-ended exploration of markets, ideas, stories, and strategies that will help you better invest both your time and your money. If you enjoy these conversations and want to go deeper, check out Colossus, our quarterly publication with in-depth profiles of the people shaping business and investing. You can find Colossus along with all of our podcasts at colossus.com. Patrick O'Shaughnessy is the CEO of Positive Sum. All opinions, Transcription by CastingWords So where to begin what will hopefully be a really fun conversation. I guess because you and I are interested in so many of the same things. I'm just curious what's on your mind today. I think we're both feeling a little frenzied and it's been going for a while. And it kind of feels like, if anything, it might get more frenzied and more chaotic. both for what you do but also the world around what you do. So just in this moment, what does it feel like? What's on your mind? An obvious thing for anyone thinking about how to navigate this period as an investor is just How does it unfurl and how is it different from the past? What do I do if I can't backtest? I was talking to a investor friend last night and the analogy he gave me was, I'm I keep saying I want to press the brakes as hard as I can, but I'm not doing it going 90 miles an hour. And I do think the question is, do you miss the opportunity on one side? Or, you know, do you make the mistake of every boom bust cycle in technology history? Especially if you like combine that with what does it mean when you're also four years into building a firm for the durability of the firm and the careers of the people? I think it's a complex question. Do you think that we... got it right in the profile that we wrote that you're sort of making a specific decision. I think we called it a wager, a bad positioning, whatever you want to call it, that there is a really important fundamental thing happening between a couple of labs. in AI and everybody else And, that we need to sort of like take up arms to make sure we don't end up in like a very monolithic... outcome? Do you think about it that way? I would answer in a maybe different way, which is, I believe in like the great man and great woman theories of history. If you have very high agency people in all of these places and they have the correct risk capital or support in the network or environment, yeah, you change outcomes. You look at very important questions of what happens to open source models and like U.S. industrial policy. If you think about the opportunity for the ecosystem in the future. Is there going to be a competitive Western open source model? Like it actually leads to like, did anybody make one? Could they like raise the money or were they willing to commit the capital and gather the talent base, build all the infrastructure and fight for the frontier or not? I definitely think that individual people and entrepreneurs can affect the outcome. I don't necessarily think of it as a war. I am. working closely with and co-invested with and have many friends at the labs. I think it's fair [5:00] To say the extreme point of view that some folks may had in and outside of those big labs of the owner of one, two, three. frontier models consumes the economy. I do not want that future. very clearly. And I don't think we are going to end up there. Do you actively want to be a great woman in this sense of the theory? No, I don't mean that from like a personal humility perspective. It's just like it's not my set of goals. I want to be the best investor in the things that I try to do. So it's not out of a lack of ambition or even confidence. It's I think some people are driven by like, I want to be the person. that like made this happen. I thought I was going to be like a software entrepreneur for the longest time. My decision that from an identity perspective, I was going to be an investor actually had a lot to do with the idea of I'm deeply curious. I like to understand things. I like to be right. I'm very motivated by working with extraordinary people. If I have a set of skills using those skills, to make them more successful, it's a pretty good fit for early stage and resting. If I want to work with the very best people and I want the companies to have impact, the firm can support a movement and support change we want to see. But I don't think it has to be me. If that's the goal, what does it take for? Do you think right now to be the best in a very competitive world? environment. Undeniably, you've done really well so far. I'm curious what it has taken to be that good and what you think it's going to take in the next year plus within Horizon to be that good. I think it's been pretty simple so far. My read of the environment was the growth of firms and generational transitions that were happening meant that it wasn't the most important. competitive landscape in early that it has been. And you have this massive technology transition happening. And if you took the bet on understanding the technology and the community and approach it from first principles, you might have better access and make better decisions than others who are less focused. All you have to do is take the risk and be focused. And then it's an execution play. That's one of those things that I think is pretty simple and is just... hard, this effort. I think today it has actually been not that complicated. It's more about what your bar is for the people you work with. And my partner, Mike, and I started with a set of pre-existing relationships and understanding. And so I think that has been useful. Surely there must be more than just outworking everyone. Mike said something interesting to me a couple weeks ago, which was there's sort of 250-ish people that he thinks about or you guys think about that are some combination of entrepreneurs and researchers and doing the most interesting things on the frontier. Yes, the people that are actually showing up in the morning and pushing this whole thing forward and that One of your goals as a firm is to be as close to those people. know them all and be as close to them and support them in as many ways as possible. I really like that idea. It's a cool idea. And that's more than just doing a bunch of meetings with people that are starting companies. From the outside looking in from the cheap seats, it looks like there's more unique stuff going on than just... The competition set was low and we focused more and we're executing better. I'm interested in the ingredients that have been so far a part of success. I won't name them, but there's one well-known LP that does the survey every year of what all the other fancy LPs, who they most want to invest with. You were either number one or two. There's something going on both with the companies you've invested in, with the performance so far, with the market perception. There's something more than just, okay, it was a moment in time. We worked really hard. I'm trying to get at those ingredients. That's why I'm pushing on it. To the point of perhaps making a bet that others wouldn't, we thought very carefully about what markets are going to matter and what might be different about the founders that we look at. look at in this era, if we are right about capability growth and the breadth of impact, that would just be... not obvious to other people. Where's the biggest difference from the status quo? How does the framework change? And I'll give you two examples. One of the things that we were really looking for in the first year was application areas, workflows and professions, tasks that we thought were a good fit for purpose for the models, which is a very technology-forward approach. Lots of people be like, "This is nonsense. You got to think about the customer problem and working from the customer back is the only way." We want to do both. And if you look at Harvey and the function of the law, rationally If you think that we can do next token prediction with language, and you knew that in late 2022, and then Law is structured language. I'm not a lawyer. But from the outside, I'm like, you need to like, read a lot of documents. And we had retrieval and you need to generate text. And there's a lot of precedent text, both inside firms and in common law and history. That feels like a really good [10:00] batch. We also took a very specific view of what is now possible and what is valuable. within what is possible and then who's aligned with us. Winston and Gabe, They believed that AI would transform the practice of the law in a very AI-pilled way. We will do enormously complex work, with lawyers. I don't know when, it could be next year or five years from now, but from the kernel of I can look at a landlord tenant agreement in California and answer a question. somewhat trivial to... I can project to doing a Activision Blizzard M&A and doing 85% of the work. That's a leap. The thing that appealed to me in that moment was the ambition of what was possible then and the technical logic of why it would work. I think that's probably a different decision-making framework from how other people were approaching it in that moment. What is that group talking about today, the 250 Researcher Frontiers group? What's the most interesting thing and the most notable difference between today and six months ago or 12 months ago or something? It is a violently competitive landscape. I think that was true 12 months ago, but even more true than it was 24 and 36. I think people are very concerned that it's a globally competitive landscape. And there's insecurity in that because I think it's a bit narrative breaking as well. I'm going to describe a belief and then question the belief. The belief is with recursive self-improvement of AI research, models that can... improve the models themselves. We are a year or two years away from some sort of exponential intelligence. I think that belief is new within the last 12 months for a lot of researchers. some driver of it that we're two years away. Andre Karpathy will actually, in a very self-aware way, say, I thought it was two years away for about 10 years. And he thinks it again, to be fair. Who can say? I do think that there is some sense of the major labs... are so compute intensive and so large from a headcount perspective now that the sense of contribution of I can move the needle. If OpenAI has 200 people, there's not that many researchers. The question of like, how do we get there is really up to every single person. Now, if the question is, well, I need $750 billion of compute spend, and we have many thousands of people working on this problem, I think people feel less ownership of the outcome. Oh, that's interesting. So it's just like a function of getting compute. That's the thing that's going to push this thing over the line, not our own human efforts. I definitely think there's a large contingent of researchers who would feel that one of two things is now true. What I do doesn't matter anyway because the model is going to do it. Or two, the only thing that matters is compute scale. And both of those are somewhat disempowering. So then what are they doing? If I believe both of those and I'm a top five researcher or something. That says something about your psychology because you want to do something that matters. I think there's also a set of scientists who want to work on the thing, whether or not it matters that they're working on it. An entrepreneur asked me yesterday, I would not work on the companies if I felt like we couldn't change the outcomes a little bit for them. He asked me yesterday, would any of the companies that you backed not been backed without you at that round? Would people have just said no? 5%, 10% max. They were resourceful, really talented people. They find other investors. There are lots of smart investors in the world who want to take that risk. maybe they wouldn't have gotten the next $100 million in compute. I don't think every researcher doing... frontier work at these top couple labs right now feels like they're essential to the machine. What do you think is unhealthiest part of everything going on right now? Like what worries you about what people are trying to accomplish and what are the things that you're most worried will inhibit the future that You want to see. I think compute is one people, I think, understand that very well. The version of us having enough compute over the next five to 10 years, and people are like very much thinking about like 2032 at this point at scale. I was talking to the leader for infrastructure at one of the hyperscalers earlier in the week. And he's like, there was nothing that is going to move the needle for us at sufficient scale before 2030. That's depressing. Where can we get sufficient natural gas? Americans and entrepreneurs have been able to build new technologies and new capabilities very, very quickly in the past. I don't think it is a technology or capability or capitalism problem. I think it is a regulatory problem and an alignment problem. And I don't mean like AI alignment. I mean, if you want to... build data centers in New York, you need to convince the people of New York they should want [15:00] data centers there, or America should want data centers there. If you want to make the price of nuclear competitive... as baseload power... then you need to convince people it's safe and you need to allow enough construction of SMRs in order for the price to come down because we understand how that cost curve will change in theory. I don't think we lack the technical and entrepreneurial capability to build abundant cheap energy for the data centers. That's one. The physical supply chain is just a tough reality. That makes me And having the tacit knowledge and the labor and the raw materials, that can't go as fast as software or even decision making. The only way through that is through. We just have to invest in it. On the investing side of what worries me, I was talking to one of our founders, several of our founders who are researchers and entrepreneurs. Explaining that the quality of their storytelling and their ability to make the companies that's legible to investors is like obviously very important to their success. That's going to be a. capex intensive play up front. The reality of the financial landscape for all of these people is I'm not a research scientist, you're not a research scientist, and all of the capital is not research scientists. It's on entrepreneurs to go explain their story. But one of the challenges is there are a lot of people attempting to invest, as they should, in technology bets. It varies how much fundamental understanding there is. There's a lot of proxying of judgment to pedigree or to other legible signals, who's invested in References are always good. Who are the referral sources? And so the decision making is less fundamental. I asked a extraordinarily good investor friend. We had a debate like you do. And I was like, I don't understand. What is this company going to be that would be big? Explain it to me. His explanation was essentially, do you know the quality of this person? Yes, I've known the quality of the person for eight years. The technical theory in the business don't are making large scale research bets without any intuition for them or without any opinion on them. And I'm like, it's not all going to work. And I may not be any better at deciding, but we want to get to that intuition. And I think not having a point of view on the business besides the pedigree of the person is dangerous. Vanta automates security and compliance for over 16,000 fast-moving companies like Ramp, Cursor, and Harvey, keeping them audit-ready around the clock. It's the number one agentic trust platform, and it now helps companies like yours watch for the risks that show up between audits, across your vendors, your AI tools, and your whole environment. Every new tool your team signs up for, every vendor that turns on AI features, is an opportunity for something to go wrong, and most security programs weren't built for AI's pace of growth. works like a 24-7 GRC engineer in the background, finding issues, drafting fixes for you, and cutting vendor assessment time by up to 50%. Whether you're a fast-growing startup or a global enterprise, Vanta helps you earn and prove trust. Invest like the best listeners get a special offer for $1,000 off at vanta.com slash invest. Ridgeline is the first end-to-end system of record with embedded AI for investment management firms, running portfolio accounting, reconciliation, reporting, trading, and compliance on one unified platform. Firms are moving off legacy technology and onto Ridgeline because of how far ahead Ridgeline's AI features are compared to anything else in investment management software. Which is why I believe that firms that come out ahead in the AI era will be the ones running on Ridgeline's unified platform. If you're serious about your firm's AI strategy, Ridgeline should be part of that conversation. You can request a demo at Ridgeline.ai. you Who is a researcher that has most blown you away and how did they do so? What's something that they did that to you felt extraordinary? Back to the great person theory. There are a number of these people who I think the history books will write about the contribution of theirs as a truly extraordinary thing that created this kink. What's an example of a person like that and something that you've seen them do? My partner Pranav and I were introduced to Tony Zhao and Chang Chi at Sunday Robotics when they were PhD students at Stanford. They worked at Toyota Research and DeepMind and Tesla. And so they weren't just academics by any means. They're young PhD students. So I think they're like 25. I think one of them didn't finish. They just like started the company. What I thought was impressive [20:00] them then and then now. I remember looking at their body of work. It took me a while to get oriented, but I think that they have contributed... Dual-handedly, most of the interesting ideas in robotics AI over the last... Four years. That's a pretty weird thing for two very young people to do. If I try to characterize the type of ideas, it is how can I use... modern AI, to solve the robotics generalization and robustness problem in a very practical way. It is believed in robotics that if we just had the Internet of robotics data, yeah, we'd have fully general robots everywhere. This is clearly going to work. One of the big research and practical problems is like, where do you get that data? Some of the ideas that Tony and Chang have worked on is... How can you be clever about collecting the data in the cheapest way possible in a way that supports the distribution of real world environments and tasks? And so I think the creativity of thinking about the actual constraints. We don't have the data. We don't have infinite dollars to spend on the data. And we're going to treat it as a technical problem to solve of the shape of the data and the collection and how to interact with the model and how much of this cheap data collection can we transfer to model learning. That's super interesting. That's very outcomes driven. I was blown away in the first meeting. We said yes immediately there. Thankfully, it's just under two years that the company has been that we are going to have general semi-humanoid robots doing things in people's homes. First and beta end of this year. That... is not something that any of us believed. It blows me away that you can move that quickly from a bunch of cardboard in a Stanford basement to the full stack thing that is manufactured here. We have done hundreds of iterations of hardware on model data collection and translated it into tasks and tested in all these real world environments. It's going to work. The videos are very cool. I think the speed of that is mind boggling. I still think the broad view, not everyone, but like lots of people in robotics are like, now it's a question of when, not if. It surprised me that the team was like, if not this year, next year. I'm very curious about the moments of your investment decisions. Like you just said, in the first meeting, you were sort of like, we're in. Is it always like that? Or are there examples of things where you like hem and haw, you end up doing it and it works? I would love to hear more about how you and your team make investment decisions, the actual in the room process for, okay, this thing is interesting, we're going to do it or we're not, we're debating it. What is that process? What is it like? It varies based on company. I'm very instinctive on people. I often know I want to do something immediately. If I know what somebody has worked on and then I interact with them and I hear the of some basis, some background in the idea. We have a rating scale, a one to 10 scale on it. I'm like immediately an eight or a nine. What I'm then doing between that and a real decision is often figuring out, What are the holes in my understanding where my judgment of their premise or them is incomplete or wrong? What do I not know? If you're an investor looking for the most ambitious, impactful companies, you can't know every domain. We have biology and defense and robotics and law. I'm like spending the next time. day to a few weeks desperately trying to ground myself and being like, okay, what does everybody else believe about this space? Are they the people I think they are? That's what my process looks like. And then I want to get feedback. I want to get like second reads on people. I want to understand what the core questions are. I'm a memo person. Even at the very beginning of the firm, when it was just me, I would write the full memo, perhaps Greylock style, ship it off to a friend that was an investor who I trusted for their perspective outside of the funds. You and I have talked about what do you value about a partnership? And I'm like, I'm comfortable making investment decisions, but I think other people can make me better. I want people to push at the logic and have me reflect. So I used to like take them out and like send it to John Lilly or Dylan Field or something. We run a memo. I go see what Bella or Pranav or Mike wants to know about the person, try to complete the picture. Other people, I think, are like more even in their decision making. They're like thinking about it and thinking about it. They climb to conviction versus I start there and then I like work backwards. But I think a similarity between both Mike and I is we'll start at a point and explain what will move us. partner Bella and I'm like okay instinctively I I like am positive on it, but I don't know enough about the science here. And I have to like go make sure this makes sense. Can you really be an eight or a nine if you don't get it? No. [25:00] I could get there. Now, as we were talking about, is one of my concerns for this period of time where I'm like, if you don't feel like you have any grounded intuition on the bet itself, what are we doing here? I'm trying to think about if that's fair. If Brett Taylor wanted to like dig in volcanoes or do dog streaming or something, I'd be like, yeah, of course, man. But he wouldn't do that. There's some version of, I don't even care if it doesn't compile in my brain. The person is so undeniably good that I would just... Back then, Brett, as an example. Yes. I feel like these things are so... in extra fully. intertwined in my mind. Because part of what makes people great, like I was trying to help one of my companies, the candidate yesterday. And they're like, Oh, what'd you see in these people? And I'm like, They're just so right. He's right all the time. Their industrial logic is impeccable. They have all these great character traits, too. They're amazing at recruiting, but they have a point of view that I think is going to be right in the world. I've seen them just make repeatedly correct decisions, including when I am wrong. And I'm like, I have a lot of respect for that. When you say like these people are amazing and Part of what I think makes them amazing is I believe in their judgment. If I don't understand what they're doing, I can't have an opinion on their judgment. Brett's doing enterprise AI. I understand what he's doing. If I were to build a pie chart of your time now, not when you started the firm, but today, there's meeting new companies, there's helping existing companies, there's talking to researchers, there's time with other people, there's talking to candidates. I have no sense of what it would be. If you had to sum it up, what are the things that you spend your time on and what's the percentage allocation? I think I spend two thirds. Three-fifths of my time working on portfolio company stuff. And that is recruiting, helping people think through things, trying to influence the outside ecosystem in some way. Then raising money. The next largest piece is... looking at companies. I don't know if this is right or wrong. I get paranoid. about it and like move the number but I probably see four to six new companies a week it's not a very high volume my first couple months like my old firm I saw 500 companies so I'm like wow now I feel more calibration where I just have much more confidence I can tell The balance of my time, I am doing a combination of help one of my partners and look at something, meet people that might teach me something about the world. And that can be a researcher. Or if you are purely early stage in a larger firm, you can be very myopic because your ecosystem is big enough. And we are very small firm or ecosystem oriented. And so I've learned so much from just getting to know investors who think differently than I do, including different asset classes. Well, I never spent that much time with public markets people before. And it was very educational how they think about the world. I spend time with other investors of all skills and asset classes. I spend time with companies right now with a pharma company who is thinking about how AI is going to transform their business. This is very interesting to me because I've learned a lot about what they believe about the future. I'm doing a lot, but I'm like leaving room in my calendar for learning, feeding curiosity. And then there are pieces that are other parts of Bridges to DC. External communication we've learned about raising money I stand with the belief that I... advise entrepreneurs with, which is you should understand people's objections to what you are doing and their questions, but you should not tell them what they want to hear. When I started... the fundraise for Fund One I knew a lot of LPs over a long period of time already, so it was not very complicated. But I had one of my friends who is a private equity investor who stated, you have to have like a very differentiated story for LPs. Every part of the funnel should be the specific thing you're going to do. And I'm like, let's be honest, I don't know yet, but I need to like raise the money so I can go experiment and figure it out. now... I think you don't know until you make contact with reality and like think about it and you're in the market. There were definitely LPs who did not like that. I gave people a two pager on my background and investing history and claimed that I was good at identifying extraordinary people. Being useful to that set of people and being genuine supporters, if you combine that with investment judgment and ability to recruit, you've got a starting point. Mostly some things I imagine about firm culture, and then we'll go execute like hell and figure it out. I can see how this is tough from an LP perspective. And I deeply value the people who bet on us early because they go write a memo for their investment committee and they're like, she's going to execute like hell. We'll find out. the cleanliness of the story is what people are looking for. I can't advise other managers [30:00] on this because people have different outcomes. If you tell people what you are going to do, life is much simpler and what you actually believe life is much simpler. And for me as a master, when somebody can convince me that the world works differently than I thought, I'm like immediately incredibly excited. Maybe I can convince people that this is just... how it actually works. I've met a lot more investment managers over the last four years. I knew a lot of BCs, but people who are doing creative things. I really like entrepreneurial investment managers, as you, I think, do. And as you might imagine, where I'm like, I don't want to build the firms that they've built. but the creativity with which... Philippe and Thomas go to or that Josh at Thrive approach their business. and the encouragement they have for others to approach their business. Well, you can do new things and you should go express your opinions in the form of your investment management firm. I think it's amazing. What was imprinted on you watching your parents who are both entrepreneurs? I think this is very... helpful to me because there was no Moment of my childhood or being a teenager where I felt like they were not there for me, even though they worked hard. All the time. Both of them. Can we try to resolve these things? I was a pretty independent kid. That helps me. I don't know where I was in the distribution, but I feel like I was pretty independent. It helps me to think your family can make you feel like you are the center of their world, but they're whole people with other interests and they want to spend time doing other things too. Even just recognize my own importance from the perspective of remembering what it's like to be 10 years old. I liked my mom. I love my mom and dad. And it was so cool. Our family values are very similar to theirs. Integrity, thinking for yourself, independence of thought. And then there's like focus on family, team spirit and family. The independence of thought is probably the least generic one of those. I think a lot of people like want to be good and kind and work hard and whatever else. For both my parents, it was a moral issue. You cannot ever worry about what other people think. I'm like far to that side of the spectrum, but I do think about it sometimes. You do sometimes worry what other people think? Yeah. What do you want them to think? That you worry that they don't? I worry about raising people's competitive values. hackles in the ecosystem. Why? Because I'm a friendly person. I want to be friends with everybody. I don't mind competition, but the sometimes pure zero sum competition stance of traditional series A firms, series A, series B firms that says I'm going to own a business. 18 to 25% of this company and take the board and you're going to own none of it. is not conducive to like a lot of collaboration. There are issues with that from an incentives perspective, but the public markets orientation, people love to tell you about their best ideas, pile in after them. This is an extraordinary situation. I would love to talk about why gaming and entertainment is going to be totally different and people are super under indexed on it. There's part of that orientation that just appeals to me as a very positive some person. I'm always interested in sources of inspiration. I'm curious in two ways, overall in your life, who has inspired you the most? And also right now in this very moment, who is inspiring you the most and why? I saw my parents build a company. I was like, this is so cool. It's us against the man. The man is very big companies. The man is trying to kill us. We can still do it just because the technology is better. The product is better and the customer will want it. My love and ethos goes to entrepreneurs. You can make something out of nothing because you see a better future and you can do it really fast. These things appeal to me in terms of what I want to try. There's so much courage in that and optimism. There is something poisonous that bothers me today where I think especially like the post Gen Z entrepreneurial crowd think it's like all marketing and brand is real. Building a network is totally real. No one denies that, but when folks are very cynical about... how the world works, how entrepreneurship works, that it's just nepotism and Twitter is useful. I think that's nonsense. If you focus on value and treating people well, and you work with extraordinary people and the vision is worthwhile, that works more times than you'd think. There's so much criticism about playing the game, be it marketing or fundraising. And I hate that. Codd is like this. I find him very inspiring. And Tuhan is like this. I find him very inspiring, right thing by the customer, we will win. Feels a lot more complicated than that. I think he's right. That seems to be working. One of the huge debates right now is what to do about the fact that there are open source models, not American made. which are competitive at the frontier of performance of AI models. And seem to clearly have been at least to some degree based on... [35:00] the work of American models. what to do about this, what it means for the future of AI. Companies love open source because they can build their own thing. And Base 10 and others that serve a lot of inference work with a lot of these models. How are you thinking about What's right? what should happen, what will happen, the implications for business, That's a big, hard, important, interesting question. It's a big question. My point of view on what is healthy for businesses, America, the ecosystem, individuals, and then there's what's already actually happened. The reality is we have, over the last three years, increasingly competitive open source models from all fronts. Largely China, but definitely also the U.S. and Europe. Most recently, Thinkie, Poolside, People Waiting for Reflection, NVIDIA Models, Mistral. We will have very powerful and already do have very powerful open source models from Western countries. The cat is out of the bag and these are in use. Everywhere even if you have no economic point of view on the labs. And you just say, how is the diffusion of capability going to happen in the economy? There are a huge number of instances where it is too expensive to too sensitive or too slow to use the model from the frontier providers today. And I think that's going to increase as we learn how to do more things with AI, because it's actually quite expensive. It's objectively true that if the capabilities are more democratized, you will see them used in more ways. I want to see that happen. I do think that It would be irresponsible not to understand the safety profile of models as they progress, because you just draw the line. We have companies that use frontier model capability for defensive cybersecurity and for biology work. If it works for those use cases, it obviously also should work in similar ways for the offensive cases. We just need to look at that reality and think about what are the other ways in which you control this. But attempting to stop technological progress and openness around it, if you did restrict use of open source models in the United States, You'd basically just restrict law abiding American businesses and slow them down or move profits, different pockets, prevent certain uses of them, because the actual attackers. or people who have adversarial uses of these things are not affected by your restrictions. You're asserting your own people. My view would be there should be testing and understanding of these models at the frontier. People are very worried about backdoor like behaviors in Chinese models. The thing to do would actually be like have a very rigorous set of safety testing on that. Like, let's go find out as much as we can instead of talking about how there might be this issue. in a speculative way where there's not been nearly enough actual research on it. The future where there is broad access to intelligence too cheap to meter, as Sam put it, that is coming. It will be supported by open source. Businesses want it to control their own destiny for economics, for capacity. If given those models and the increasing democratization of the skills to post-train these models, build harnesses, use tasks, the economy is so big. imagined by a researcher in a frontier lab. You can't imagine the diversity of reality. Even if you trusted the models to go... figure out what to do. They have to get there. The best way for that to happen is a ecosystem of businesses, as we've always had in the economy, cheap infrastructure. Do you ever worry that an alternate version of U.S. history is that there was energy too cheap to meter because we built 1,000 AP1000s or something? much like China is doing now in nuclear. And just a set of circumstances happened such that we just didn't get that. Do you ever worry about that as it relates to intelligence? It does seem inevitable that we're going to have abundant, accessible, low-cost, valuable intelligence. Can you imagine a world where we don't? Yes. Absolutely. And I also think I can imagine very easily a world where we don't have that in a competitive way because it is essential to... Economic competitiveness and national security. There is not a version of the world where we rebuild our industrial base without automation in the United States. If we don't import people and our people are expensive and we lack some of the skills, but want to produce a lot more goods and have a more resilient supply chain. Just doesn't add up. Who's going to produce this stuff? [40:00] Well, the United States do not want to work. and should not want to work for $13 an hour doing a very inhuman job. I don't think it is inevitable that we are competitive. And I think we need to make that decision actively. The version of it that I think is very... possible is that people are rationally afraid of the impact of AI on jobs or dislike the capture of rent by a small number of technology firms, rejecting the idea of being in the permanent underclass and then connecting that to a anti-capitalist. orientation, that contingent of thought can slow down build of energy and infrastructure and and industrial capacity. One of the most important inputs is compute. If we don't have it, we're naturally not competitive, or we're at least not independent. I think we're going to start talking much more about compute independence. That's a big problem. So on this point of compute independence, what does that mean? What are the missing pieces of compute independence? Like the most obvious one might be more fab capacity, more leading edge fab capacity here in the U.S. or something like this. There's all sorts of stuff upstream. There's particular kinds of glass, very important. It's controlled by basically one company that TSMC has like a monopoly on the supply of. There's all these component parts. But if you think about compute independence, like if so much boils down to compute, what's to be done about it? Are you trying to invest in companies Solving that problem. What is the problem? Say a bit more about what it'll take. Well, if you just work backwards from... A data center full of GPUs, cooling, powering, training and inference so we can support the use cases. All of those inputs, it actually looks a great deal like energy independence or something like that. All those inputs, there is a global supply chain. I've got my TSMC mug with me. There are parts of that supply chain that are like a very thin sieve in a place that is not necessarily stable or accessible to the U.S. and allies. A different version of the world that will take a bunch of investment and national security and energy policy is, for example, Jacob Halberg is working on something called Paxilica. And it's like, OK, for every part of the supply chain, can we invest in more capacity and figure out what the independent paths are? I don't think that means it's all got to be created in the United States. Comparative advantage is real, but having more than one source is a position that everybody wants to be in. When we think about the components that we've invested in, we've invested in the labor gap for data centers and robotics. We've invested in nuclear energy. We've invested in alternative chip architectures. We keep looking at people who are essentially... data center builders, solar and battery installers of some kind. Part of this is the actual capacity buildup. That's a very operational business, like somewhere between operations, some technology and real estate. And financing. And financing. Absolutely. That's probably the dominant thing. We have not invested in it yet, despite looking very closely. I'm not opposed to it, but fundamentally, I'm a technology investor. I want to understand what it is that is the durable product asset people are building. What are the base debates inside of Conviction? You've person, amazing engineer. The young talent in your firm brings really interesting perspectives. So I'm imagining great lively debates about things that matter, et cetera. What are the big debates today? Alotta and I have a podcast called No Priors. This premise of some of the things you believed, especially about markets as a pastor, like no longer true, is an interesting one. And regularly, we look at companies in domains that are not traditional software domains. Not even traditional software domains, but can you make money in this market at all? Venture investing in semi's companies, Liputan aside, was like a god-awful business for the longest time. The returns... Everyone told me this. Yes. It was really bad. That's like an example of Bella, a partner on our team, started looking at a bunch of these companies and it's obvious that the demand is there. Now I think we have arrived at the conclusion that others are as well, which is... Well, the market is different today. we are seeing consolidated at scale demand for people accelerators or even supply chain independence because the big buyers of it want it too. We can't all be stuck on one line at GSMC. All of that is very, very valuable. And that changes the risk equation for these companies. We start with pretty aligned beliefs about the direction of travel or the problems that are worth working on. And then a lot of the debates are like, is the market friendly to a venture project? [45:00] company or not. Space is not a friendly market to a venture-backed company, but is a possible market. Is the distribution of outcomes worth betting on? That could be true in solar and batteries and nuclear in turbine manufacturing in robotics and biology. These are not your favorite software markets from 10 years ago. Each of them is a new debate. Biology is an interesting one where by virtue of seeing the data empirically, I have now strongly moved to one side of the debate. What side is that? You can create and capture enormous value with models in biology. And there could be different AI software in biology. We're the first check in a company called Chai Discovery. And Chai is working with a number of top 10 pharma companies in really significant ways to accelerate some part of the R&D process. The conventional wisdom when we invest in this company, and I've been looking at computational biology companies of different sorts for five plus years at that point. The conventional wisdom was the only way you make money in biotech or serving pharma is by make a drug and then get biobux deals and then decide how far along that risk path you want to take. And what that means for the capital structure of the company is... OK, the great traditional biotech firms find these principal investigators and they own 40 percent of the company and they're kind of assembling these things. turning them into candidates, but most of it doesn't work. So it's just like a very different distribution of outcomes and structure and way to invest in businesses. Dumb software investors, you can't make money selling software to pharma or build platform businesses in pharma. The debate is, does it change with models? I'm a strong yes. Now, we still have a question to solve on regulatory issues. There's the speed of the physical world and safety is not something you can overcome easily. But I think we should see a massive acceleration in cures. What flipped that for you? What evidence did you see from we don't know yet? We invested when we didn't know yet. As part of the fun adventure, we were like, it is possible and it is worth trying. It could happen, yeah. There's no genius here. That's a $10 million contract. The other piece is actually you just talk to the scientist at the customer or somebody who leads a division. The user of the tool. Yeah, the user of the tool. Or you see there are also end user adopted tools, product-led growth tools that are working in the space. Well, the customer knows if it's valuable or not. light bulb moment for the industry is when we will have a new indication or a new drug that clearly the trajectory of the thing was changed. created by AI. We should see a huge wave of investment and rightfully so, but it's going to happen. I'm like very impressed by the speed with which pharma and healthcare overall has said, yes, this is going to make a difference. And we actually think it's going to change the business. Why do you call it conviction? It's aspirational. The most traditional form of early stage investing is you start early with a company. You have a significant position. You never sell the position and you work on the company until it works or is sold or dies. There's wonderful alignment and simplicity to that. Mike and I have both had the benefit of being part of the journey for some companies where it Took a minute to begin to work or it's like not obvious at the beginning. Sigma, notion, rippling. You wouldn't bet on a company hoping that it's going to take four or five years to find the thing. I think everybody is a product of their own experiences, including investing experiences. The first couple of years at base 10 were very non-obvious as well. the ability to take a point of view that is not obvious in the market, because the people's backgrounds, whatever reason, and then just suspend doubt and act with full belief until it's true or not. That's a great way to be a partner to somebody building a business. I am trying to build a partnership and I deeply believe this is a team sport. I was talking to a friend who runs another investing firm where he is like, we don't have individual ownership of our investments. And I'm like, that is nonsense to me. How could you run a business that way? Because somebody has to this is right or wrong, but I don't know any other way to invest by saying, "Oh, Patrick, you must make the decision. Do you believe, convince us, how can we help you make that decision? What have you learned about risk-taking? Behind conviction, it sounds like there is often a leap of faith of some sort. Obviously, if you knew everything and it was obvious, that would be priced in and there'd be no return opportunity. Is there always a leap of faith? Is that risk-taking by another name? Unlike my good friends at Founders Fund, I don't have an instinct to be contrarian, but I do think it is so fundamental to decide what you think and not worry too much about what other people think. Other people is... the [50:00] dominant narratives of the period, or even what different players in the ecosystem that are really important declare one way or another. I think you just need to find the truth. The way I would relate that to risk-taking is if you find the truth and it is wrongly priced and you hold on to that, you're in a good position. You want asymmetric information and then the confidence to hold the opinion when other people haven't come around to it yet. I think a lot about how to make sure we have the information that is better than other people's and then protect ourselves from Noise. Describe that in one more level of detail. That's a great way of asking like what conviction looks like today. You're building that shell. What are the keys to doing those two things? Well, both having the truth and protecting yourself from the noise. I like want to spend my... time when I'm learning about the world from somebody who... is making it happen. Our portfolio founders or the many founders who are doing amazing things outside of portfolio, who are doing something that surprises me or advances their frontier or like really smart believe something I don't. These are like three different categories. And I'll give you an example. Mikey Shulman at Suno is building an amazing business doing music generation. Shame on me. I knew Mikey, a mutual friend of ours who's an investor, like asked me to like invest and I stupidly said no. But I was just like, oh, I don't think that many people want to make music. I make music. But can you turn into a social network? How much consumption is there going to be? A lot of questions. My intuition is just wrong. the amount of expression or entertainment and creation for a lot of AI tools. I've learned something here by talking to Mikey about his business and what people are trying to do. So if it is founders, who have companies that are creating a behavior you don't understand, somebody working on research and an interesting direction, businesses that are like, here's my plan for AI. All that is super educational. I think the circular logic sometimes of what do people believe about the big lab strategy today? How can any of the applications live? It's actually not that instructive for your decision making. safe way where they capture a lot of profit. The priorities that ladder into that probably look like chat GPT, ads, coding. Maybe there's an expansion after that. co-work, the ability to get different types of users to do richer tasks and more interfaces. I think you have to judge the actual competitiveness of any of those efforts, the reasonable scope of them, and then look at them relative to each of our companies or opportunities that we're looking at. Coming up with some grand strategic framework for like What layer is going to win here, I think is not useful to me. And I feel like people spend so much of their investing energy thinking about that. Versus I want to spend my energy at figuring out if we're 1% of the way in what is the next 99% of diffusion? That's where we try to direct our energy. For fun, as we wind down here, understanding this is like a purely speculative question, and it's meant more fun than raw prediction. What are some things you think are true a year from now, based on all these incredible people that you're close with, the research community, the... entrepreneurs like the Sunday founders, you add it all up. Things are moving fast. A year is a long time and it's like reverse dog years now. What do you think is notably different about the world of technology a year from now? I'm hopeful that a year from now, we see Jevin's paradox in practice as we have a Agents and products that do more of the mundane more effectively in all the domains of our lives, it should look like the transformation that has happened in software engineering. You have companies, I have companies where they're like, we're just going way faster. I expect that some analogy like that. will happen everywhere else. And we see it in our companies where I'm thinking about in one of our portfolio companies, the marketing department. It's like a person in the house. This is a company that serves lots of customers. They need to do very traditional things, sales enablement content. What happened was the guy in charge of marketing is interested in creating leverage for himself. And he's like, I made an autonomous marketing department for us, the company. In every function, as you learn faster, do less of the mundane, you will repurpose that time somehow. Do you work less now that you are more productive with AI? Yeah, I work more. And I think this is like a core wisdom of Jensen. [55:00] which is we're all going to be more employed and we need to make sure that people are given access and education to the tooling that will allow that to happen. But what you built is incredible. We've loved through the Colossus side, getting to know your whole world. It's so distinctive. You are a great example that there's always room for great, meaning there was plenty of early stage investment firms when you started conviction. And yet here we are four years later or whatever. And if you ask the people you've worked with, you've made a real huge difference in their lives. There's just always room for great. I think that's like a great, awesome lesson, especially because I love how you described the early pitch was not like, here's how we're differentiated at every level of the funnel. thing. By the end, I got so frustrated. I was just like, it's an execution game. Totally. I ask everyone the same traditional closing question. What is the kindest thing that anyone's ever done for you? I love this question. I'm going to give like a collective answer. I think there are so many people who are... extraordinarily accomplished in Silicon Valley. who care very little for pedigree. As soon as they have a conversation with you, and they're like, maybe you can help me, or you have an interesting idea, or... Maybe I just think you're promising. And the dominant factor in their own willingness to invest in a relationship where a person is just their assessment of the idea and the person. I think that's amazing. That is not how most ecosystems work. And so Ashim Channa and Anil Bussery and Joseph Ansonelli and Reid Hoffman, who hired me at Greylock, I started when I was 23. People love to make fun of young VCs where they're like, ah, what a barnacle on the ecosystem. It's a terrible experience for entrepreneurs. I don't know anything. They're trying to advise people who gave this kid money. And I'm like, Well, one, my job is just to make other people successful at the time. You can take any task in any job and just try to be great at a task with mimicry and first principles thinking. We hire earlier career people at my firm. But I do think, oh, my goodness. Thank you for taking a shot on... some random person and then investing the time to like teach me how to be an investor. I think there were a few people who gave me advice. starting the firm who I think they would think of this as entirely trivial. I'm just giving you my opinion. But Ravi Gupta, who is now co-seer of a new thing called Ithaca, Dylan Field and Elena Natalinsky, John Lilly, who's been a longtime partner and friend. And There are a few folks who were just like, you can definitely do it. I was going to do it either way. But having the encouragement of people who believed that there was room to be great, and including some of our first LPs, I will be forever grateful to, I guess, the people who took risk with me. It's beautiful the world runs on faith, belief without evidence yet, and still conviction in someone's ability to do something. Pretty cool. It's faith in people. We talked a lot about how people's ideas and our opinions of them are intertwined. But I think that's a beautiful thing because you don't need any particular advantage to have an idea. The fact that and put faith in us, I could not be more grateful. I've learned a lot watching you operate and talking to you. This has been really fun. Thanks for having me. Thanks. If you enjoyed this episode, visit Colossus.com. You'll find every episode of this podcast complete with hand edited transcripts. You can also subscribe to Colossus, our quarterly print, digital and private audio publication featuring in-depth profiles of the founders, investors and companies that we admire most. Learn more at Colossus.com slash subscribe. Bye. Outro Music. You know how small advantages compound over time? That's true in investing and just as true in how you run your company. Your spending system is your capital allocation strategy. Ramp makes it smarter by default. Better data, better decisions, better economics over time. See how at ramp.com slash invest. As your business grows, Vanta scales with you, automating compliance and giving you a single source of truth for security and risk. Learn more at vanta.com slash invest. The best AI and software companies from open AI to cursor to perplexity use WorkOS to become enterprise not in months. Visit WorkOS.com to skip the unglamorous infrastructure work and focus on your product. Ridgeline is redefining asset management technology as a true partner, not just a software vendor. They've helped firms 5x in scale, enabling faster growth, smarter operations, and a competitive edge. Visit RidgelineApps.com to see what they can unlock for your firm. Every investment firm is unique and generic AI doesn't understand your process. Rogo does. It's an AI platform built specifically for Wall Street, connected to your real outputs. Check them out at rogo.ai.com. [60:00] invest.