Replit CEO Masad Warns Young Founders Against Rushing to Raise VC Cash

Replit CEO Masad 告诫年轻创业者切勿急于寻求风投 在真正的求知欲指引出毕生事业之前，十八岁创办公司会让创造者过早陷入长达十年的公司承诺和股权结构表义务中。 Masad 与 Biyani 认为，传统大学更看重服从、标准化评分和文凭主义，而非自发的求知欲，从而扼杀了社会对思想异见者的需求。雄心勃勃的年轻创造者应当避开过早优化的陷阱——例如在十八岁时就背负长达十年的风投承诺——转而在 AI 的辅助下展开低风险、基于项目的探索。演讲者结合自身经历（例如 Amjad 偶然钻研国际象棋从而催生了 Replit 的自动研究机器学习项目），指出未设限的痴迷能够带来非同寻常的现实突破。至关重要的是，他们坚称技术上的逆向思维必须与心智成熟、亲社会美德以及高度信任的协作相结合，从而彻底终结业界过去对社交功能失常型天才的推崇。 人工智能消除了大众个性化教育的执行障碍，使基于项目的学习和自主探索成为切实可行的常态，而非边缘案例。 (11:47) 催促18岁的年轻人去创立风投支持的公司往往构成一种“过早优化”，在自发信念形成之前，就让杰出人才背负上长达十年的义务和股权结构表负担。 (14:46) 自主 AI Agent 可能会在无形中化解管理内耗，从而复兴像合弄制（Holacracy）这类此前失败的组织形式实验，推动“自动驾驶公司”的出现。 (24:02) 硅谷正在主动摒弃社交功能失调的“自闭天才”陈旧形象，转而重新高度重视亚里士多德美德伦理、心智成熟度以及公共沟通能力。 (36:38)

## Background 本期内容由风投机构 a16z 于 2026 年 9 月 23 日发布在 YouTube 频道，由 Erik Torenberg 主持。对话嘉宾包括 Replit 创始人兼 CEO Amjad Masad，以及 Horowitz and Andreessen Academy 联合创始人兼 CEO Gagan Biyani。节目简介指出，本次访谈聚焦于 AI 时代成长起来的年轻一代需要何种教育，探讨了年轻人挑战成见成为“异端”的社会价值、通过项目驱动与“支线任务”重构学习体系，并反思了年轻创业者过早专业化与被资本绑架的陷阱，以及个人信任与品德维度的发展。 ## TL;DR Amjad Masad 与 Gagan Biyani 直指传统高等教育的规训弊端，呼吁大学应保护年轻人的叛逆直觉，借助 AI 转向完全由兴趣与实践驱动的“项目驱动学习”并包容各类“支线任务”。两人同时警告切勿在 18 岁盲目创业而陷入“过早优化”与掠夺性资本的桎梏，并以从研究国际象棋自发演进到自动化机器学习研发为例，阐释了好奇心自我滋养的递归闭环。此外，对话展望了 AI 赋能的“自动驾驶公司”组织演变与硅谷的高信任杠杆，强调在不可预测的奇点时代，青年更需借助亚里士多德美德伦理学锚定道德底线，兼具硬核钻研的“极客执念”与融入社会的谦逊协作心智。 ## Atmosphere & style 整场对话洋溢着硅谷精英智识圈层特有的锐利锋芒与务实黑客底色，在辛辣解构传统教育体制的冷峻反思与展望 AI 赋能未来的蓬勃雄心之间自由穿梭。Amjad 展现出罕见的双重特质，既有动手写代码训练模型的极客穿透力，又兼具博览文史经典的哲学沉思，与 Gagan 充满实战细节的教育实验观察形成强烈共鸣。聆听过程宛如一场智识冲浪，既有第一性原理推翻教条的解构快意，又贯穿着对品格成熟、信任沉淀与公共责任的深厚温情，带给听众兼具宏大格局与微观反省的认知震颤。 ## Bulletin - [11:55](https://www.youtube.com/watch?v=Q8yFrGwIftE&t=715s) 随着 AI 具备自主执行机器学习研究的能力，高等教育亟需从指令性灌输转向兴趣主导的“项目驱动学习”，大学角色应转变为向学生的自发探索提供顶级算力与资源的加速器。 - [14:46](https://www.youtube.com/watch?v=Q8yFrGwIftE&t=886s) 硅谷风投对大学在校生的“预想法”（pre-idea）百万美元注资加剧了创业的“过早优化”，许多年轻人因同侪嫉妒在未形成真正使命前仓促绑定股权，过早扼杀了自由试错与智识成长的时间窗口。 - [24:02](https://www.youtube.com/watch?v=Q8yFrGwIftE&t=1442s) AI 正在成为重塑企业架构的底层黏合剂，通过自动化承揽后台官僚流程让“自动驾驶公司”（self-driving company）成为可能，赋予员工更彻底的自主权与敏捷度。 - [28:46](https://www.youtube.com/watch?v=Q8yFrGwIftE&t=1726s) 硅谷创投生态展现出极高的信任溢价，口头投资意向兑现率达 75% 至 100%，远高于地方政治生态的 20% 至 50%，这种高信任机制显著降低了交易与沟通摩擦，应成为新型教育赋能年轻人的范本。 ## Key takeaways - [00:00](https://www.youtube.com/watch?v=Q8yFrGwIftE&t=0s) 年轻人在人类文明中最不可替代的价值是成为挑战传统教条的“异端”，而现行大学体系往往将其规训为墨守成规的平庸工具人。 - [05:36](https://www.youtube.com/watch?v=Q8yFrGwIftE&t=336s) 理想的高等教育应彻底摒弃僵化考勤与指令性教学，效仿乔布斯旁听书法课的非线性路径，在年轻人陷入狂热构建时提供充分的时间庇护与算力资源支持。 - [11:55](https://www.youtube.com/watch?v=Q8yFrGwIftE&t=715s) AI 工具的普及使大规模个性化教学成为现实，教育应回归以具体目标和真实产品为核心的“项目驱动学习”，让学生在实践中自主掌握技能。 - [14:46](https://www.youtube.com/watch?v=Q8yFrGwIftE&t=886s) 盲目过早创办初创公司属于人生的“过早优化”，许多创业动力源自同侪嫉妒而非真实好奇心，容易使年轻人在缺乏沉淀的情况下过早锁死发展路径。 - [18:11](https://www.youtube.com/watch?v=Q8yFrGwIftE&t=1091s) 风险投资机构对高校学生的掠夺性投资（如对模糊甚至尚未成型的想法开出百万美元支票）会过早固化股权结构，其负面效应堪比传统唱片业的苛刻合约。 - [20:19](https://www.youtube.com/watch?v=Q8yFrGwIftE&t=1219s) 纯粹好奇心驱动的智识探索与动手构建能形成闭环：对下国际象棋的业余钻研与模型训练，直接启发并催生了 Replit 的自动化机器学习研究新产品线。 - [24:02](https://www.youtube.com/watch?v=Q8yFrGwIftE&t=1442s) AI 智能体正接管繁琐的行政流程，以往失败的去管理层实验有望在 AI 赋能下演变为自我运转的“自动驾驶公司”。 - [26:30](https://www.youtube.com/watch?v=Q8yFrGwIftE&t=1590s) 创作者应当敬重前人建立的工业与技术基石，但绝不能迷信权威，必须拥有大刀阔斧推翻旧规则、改写历史的逆向叛逆心境。 - [28:46](https://www.youtube.com/watch?v=Q8yFrGwIftE&t=1726s) 硅谷高信任生态带来的极低沟通内耗与超高兑现效率是其创新的核心引擎，新型教育模式应将这种信任下放给探索未知边界的学生。 - [31:51](https://www.youtube.com/watch?v=Q8yFrGwIftE&t=1911s) 具备黑客倾向与边界探索冲动的青少年不应被简单定罪惩戒，而应通过合法沙盒、CTF 攻防竞赛和漏洞悬赏机制被正向疏导至网络安全领域。 - [35:40](https://www.youtube.com/watch?v=Q8yFrGwIftE&t=2140s) 判断力（discernment）源自对“非对称收益”与“非对称下行风险”的权衡，唯有通过密集承担真实决策与试错演练才能有效建立。 - [36:38](https://www.youtube.com/watch?v=Q8yFrGwIftE&t=2198s) 科技生态已融入社会建制主体，从业者必须告别“自闭天才”的人设包装，培养能够面向公共领域与政策制定者有效沟通的健全人格。 - [41:06](https://www.youtube.com/watch?v=Q8yFrGwIftE&t=2466s) 在技术奇点临近导致未来难以预测的背景下，基于后果预测的功利主义已然失效，年轻人更需要依靠亚里士多德的美德伦理学与极简公理准则来指导行为。 - [44:02](https://www.youtube.com/watch?v=Q8yFrGwIftE&t=2642s) 最具潜力的年轻人才既需要在细分领域展现出极深的研究穿透力（spike），又必须保持谦逊并具备良好的人际交往能力，避免陷入目中无人的自大误区。 ## Detailed walkthrough ### [00:00](https://www.youtube.com/watch?v=Q8yFrGwIftE&t=0s) 《城市与群星》的隐喻：年轻人的终极价值是成为“异端” 访谈伊始，Amjad Masad 抛出了一个直击本质的核心论点：年轻人在人类社会中最不可替代的效用，绝非按部就班地接受规训，而是去挑战那些根深蒂固的既定信念，充当推动文明突破的“异端”（heretics）。然而，当今绝大多数大学教育所做的恰恰相反——它们像是一台精密的模具，无情地碾碎任何出格的念头，傲慢地告诉学生“我们已经摸清了一切运转法则，你只需按部就班学习规矩，成为合格公民，然后找份工作、结婚买房买车过完一生”。但 Amjad 直言，这种传统叙事早已彻底破产：现实是许多年轻人即便在名校名列前茅，走出校门依然身负巨额债务、面临毕业即失业的窘境。 为了具象化这种文明级的错失，Amjad 引用了科幻大师阿瑟·克拉克（Arthur C. Clarke）的经典小说《城市与群星》（*The City and the Stars*）。故事构想了一个被通用人工智能（AGI）全面接管的未来都市：技术解决了战争、疾病与匮乏，人类获得了永生，每天沉迷于长达千年的艺术项目等无尽的琐碎娱乐中以消磨时光。为了应对无限寿命带来的海量记忆负担，居民会定期前往实验室，将自身记忆进行类似今天大语言模型所谓的“上下文压缩”（context compression），随后进入重训营重新“诞生”为同一个人。然而，系统偶发的一次故障导致了一个完全真正新生的个体降生。这个纯粹的新人从训练营开始就行为乖僻、打破常规，甚至开始追问所有人避而不谈的禁忌话题：“这道围墙之外是什么？群星之间究竟隐藏着什么？”周围的居民惊慌失措地警告他“外面很危险”，这种对未知的恐惧与盲从，与科幻作品《地堡》（*Silo*）和《辐射》（*Fallout*）中的设定如出一辙。 Amjad 指出，即便在被视作全球创新策源地的硅谷，本质上也是一种存在严重认知盲区的“单一文化”（monoculture），许多颠覆性想法恰恰不可能在硅谷的思维定势中自发产生。社会若要获得跃迁，就必须依赖那些拥有禅宗心智（Zen mind）与第一性原理思维的年轻人。大学最根本的使命，理应是给这些尚未被现实偏见污染的年轻人提供试验场，鼓励他们尝试常人眼中的“不可能”，从而抓住那些因底层技术范式突变而刚刚诞生的历史级机遇。 ```mermaid flowchart TD accTitle: 传统高等教育与异端创新模式的机制对比 accDescr: 对比传统大学与理想创新环境在面对年轻叛逆思想时的不同机制与产出结果。 subgraph Traditional["传统大学模式（同质化规训）"] A["学生产生'离经叛道'想法"] --> B["灌输既定知识框架与合规要求"] B --> C["强制打卡考勤 / 挤压探索时间"] C --> D["陷入单一文化与职业内卷"] end subgraph Heretic["理想培养模式（异端创新机制）"] E["产生'第一性原理'直觉"] --> F["宽容并支持学术与实践'支线任务'"] F --> G["主动倾斜学校算力、实验室等资源"] G --> H["促成范式转移与突破性技术创新"] end ``` --- ### [05:36](https://www.youtube.com/watch?v=Q8yFrGwIftE&t=336s) 重新构想 18 至 22 岁：摆脱指令性教学与拥抱“支线任务” 面对主持人关于“如果让你在 2026 年重活一次 18 到 22 岁，你会如何设计理想大学体验”的提问，Amjad 提炼出了重塑高等教育的三大支柱原则。首要原则便是彻底摒弃“指令性教学”（non-prescriptive learning），赋予学生自由漫游知识荒原的宽容度。他举了史蒂夫·乔布斯（Steve Jobs）在斯坦福毕业典礼演讲中“连点成线”（connecting the dots）的著名案例：乔布斯当年因为交不起里德学院（Reed College）的高昂学费而退学，但他选择留在学校“蹭课”，凭着纯粹的直觉去听了一门看似毫无商业价值的书法（calligraphy）课；正是这种漫无目的的知识偏航，在十年后直接启发了初代 Macintosh 电脑设计出人类历史上首个具有优雅比例间距的优美图形界面字体。Amjad 感叹，现代功利社会往往将这种非线性的闲逛斥为无用，殊不知这往往才是大学能够给人生带来的最深远转折点。 第二个原则，是当年轻人陷入“狂热建造”（manic building）的状态时，制度必须给予绝对的时间与空间庇护。Amjad 回忆起自己的大学时光，学校有着死板严苛的考勤制度，一旦缺席六次某门课就会被直接取消资格。然而对真正的创造者而言，当一个令人彻夜难眠的绝妙点子迸发时，他们需要的是能够全力以赴冲刺的土壤，而不是因为缺席几堂照本宣科的讲座而面临退学处分。第三个原则则是同侪质量的筛选：长期与墨守成规的人混在一起，再有天赋的人最终也会沦为平庸的规矩遵从者；唯有身处一个全力支持质疑权威的群体中，真正的创造力才能生根发芽。 Gagan Biyani 对此深有共鸣，并分享了自己在加州大学伯克利分校（UC Berkeley）求学时的亲身经历。他当时最大的幸运在于大部分课程并不强制抓考勤，因此他曾在一个学期里同时兼顾三份工作。他的日常操作是在学期初拿到教学大纲后，直接锁定两场期中考试的时间，向公司老板请假两天去应试，其余时间则全身心投入到真实世界的工作实践中。Gagan 观察到，这种在二十年前被视作奇葩异类的特立独行，如今已然成为许多顶尖大学生的普遍常态；但荒谬之处在于，现代大学的底层制度非但没有对这种现实提供系统性扶持，反而依旧将其视作一种需要被遮掩、违规打擦边球的灰色行为。对话双方最后强调，一个健全的教育机构绝不该让学生在追逐“支线任务”（side quests）时逆流抗争，反而应该在发现年轻人萌生出爆发式火花时，主动将高校的顶级算力、生物实验室和科研资源源源不断地输送给他们。 - **乔布斯书法课案例**：在里德学院退学后选修的书法课，成为十年后 Macintosh 首创可变间距字体与图形界面排版的决定性灵感源头。 - **僵化考勤的弊端**：传统大学常设严苛的缺课红线（例如缺席 6 次即取消修读资格），直接扼杀了学生在灵感爆发期沉浸式构建原型的可能性。 - **伯克利实战经验**：Gagan 在 UC Berkeley 期间曾单学期同时打三份工，仅在期中考当天返校应试，印证了脱离被动课堂、主动投身真实项目的巨大价值。 - **高校角色的转变**：未来的大学应从“按部就班的看门人”转变为“资源加速器”，为学生看似异想天开的突破性项目提供算力、硬件与实验室支持。 ### [10:40](https://www.youtube.com/watch?v=Q8yFrGwIftE&t=640s) 从被动灌输到以兴趣为导向：重塑实用教育 谈话延续了上一节对传统大学与学生创新行为之间脱节的讨论。Gagan Biyani 提到，有些学生在学校里甚至不得不躲进洗手间接听投资人的电话，因为学校的环境对创业极不友好。正常的教育机构本该顺水推舟，为学生提供辅导、对接资源或组织 Demo Day，而不是迫使他们四处躲藏。随后，讨论延伸到了兴趣与学习动机的本质。就像当年乔布斯去旁听看似随机的书法课一样，很多看似无关的兴趣未来都可能产生意想不到的联系。 Gagan 分享了自己童年的亲身体验：他小时候痴迷于篮球，但也常常在数学课上质问老师“学这些到底有什么用？”。如果当年的老师能告诉他，现代篮球已经高度依赖数据分析，了解薪资帽（salary cap）和高阶数据指标才能成为优秀的“比赛研习者”（student of the game），他就会立刻对数学产生极大热情。这也正是他对 Alpha School 等新型教育模式感到兴奋的原因——它们懂得如何从学生的天然兴趣出发，自然而然地引导其掌握未来所需的实用知识。 ### [11:55](https://www.youtube.com/watch?v=Q8yFrGwIftE&t=715s) AI 时代的定制化学习与“项目驱动学习” 虽然教育学术界很早就达成了共识，承认如果能为每个学生量身定制课程体系是极具价值的（正如 Alpha School 的 Joe Lonsdale/Joe Lemon 常强调的那样），但在过去，大规模定制在现实操作中几乎是不可能的任务。然而，当下的 AI 革命彻底扭转了这一局面——AI 让定制化不仅可行，甚至让传统的“一刀切”教学模式显得难以理解。以 Horowitz and Andreessen Academy（HAA）为例，如果学生要学编程，学校绝不该仅仅让他们上一门枯燥的 Python 基础课，而是应当鼓励他们直接用 Python 去构建一个自己真正渴望做出来的项目。在 AI 工具的加持下，学生完全具备自主探索的能力，不再需要老师手把手制定详细的教学大纲或课后习题，他们最需要的只是“去做这件事的自由与信心”。 Amjad Masad 犀利地指出，“项目驱动学习”（Project-based learning）在今天被当成一个时髦词汇本身就很荒谬，因为这本应是教育的默认常态。不知道所学知识的实际效用，纯粹是一种现代社会的畸形发明。例如在统计学课堂上，学生却从未被告知这些工具可以用来分析篮球或是支持自己的其他爱好，这种脱离现实应用的教学倒置存在已久。Amjad 坦言自己是个极其实用主义的人，纯粹对知识的抽象追求并不能驱动他，唯有具体的技能——那些能帮助他打造产品、实现梦想的技能——才能激发他的学习动力。只要偶尔遇到一位能讲清知识所对应技能的优秀老师，学生的潜力就会被彻底点燃。 ```mermaid flowchart TD accTitle: 传统填鸭教学与 AI 时代项目驱动教学对比 accDescr: 对比传统教育脱离实际的路径与 AI 支持下由兴趣和技能驱动的构建路径 subgraph Traditional["传统教育模式"] A1["抽象理论与无关联课程"] --> A2["强制作业与标准化测试"] A2 --> A3["学生迷茫：不知道学了有何用"] end subgraph Modern["AI 时代项目驱动模式"] B1["个人天然兴趣（如篮球、做游戏）"] --> B2["确立具体构建目标"] B2 --> B3["利用 AI 与工具自主攻克技能"] B3 --> B4["完成真实项目并掌握核心知识"] end ``` ### [14:46](https://www.youtube.com/watch?v=Q8yFrGwIftE&t=886s) 过度专业化与“过早优化”的创业陷阱 Erik Torenberg 提出了一个针锋相对的反思：虽然传统教育缺乏实用技能培养，但如今也存在另一种极端批判，即年轻人是否在“过早专业化”，以至于过早丧失了童年或缺乏智识与人性的全面滋养？传统大学虽然低效，但确实提供了读伟大经典著作、结交朋友、社交体验等非技术层面的熏陶。面对这个矛盾，Gagan 认为当前的年轻人面临着残酷而狭隘的“二选一”：要么进入四年制大学忍受极其僵化的课程、考试与绩点约束；要么直接跳出来创业，次日就能拿到 5 万美元的天使投资。事实上，绝大多数 17、18 岁的年轻人即便渴望经商，也迫切需要两到三年的缓冲期去自由探索、阅读和试错，而不是过早背负长达十年的创业重担。非营利性 Grants 或短期探索项目值得提倡，但过早拿 VC 的钱并不适合大多数普通少年。 Amjad 借用软件工程中的核心概念发出了深刻警告：“过早优化是万恶之源”（Premature optimization is the root of all evil）。他认为盲目开启初创公司本质上就是一种过早优化。如今硅谷存在一种不良风气，许多年轻人看到同龄人融巨资、跻身独角兽，其创业动力完全被同侪嫉妒（envy）所绑架。为了攀比，他们在缺乏深思熟虑的情况下仓促敲定一个自己根本不关心的项目，提前将人生锁定在一条尚未看清的道路上。 ### [18:11](https://www.youtube.com/watch?v=Q8yFrGwIftE&t=1091s) 掠夺性投资与创业的“重力法则” Amjad 强调，唯有基于真正好奇心驱动的想法才经得起时间的检验。当一个人遇到那个真正愿意倾注未来十年甚至一生的使命时，那个想法本身会产生巨大的“引力”，让人根本无法抗拒。以 Replit 为例，Amjad 回忆自己早期甚至极力克制不去创办它，但项目的巨大动量与自发扩张最终推着他走向了创业。那种顺应势能、“不得不做”的体验，才是创业应有的健康状态。 相反，当前资本市场的某些做法甚至演变成了掠夺性行为（predatory）。据 Amjad 了解，甚至有 VC 跑到斯坦福大学校园，在学生只有模糊概念甚至“完全没有具体想法”的情况下，直接开出一张 100 万美元的支票。一旦签下协议，学生还没开始探索，股权结构表（cap table）就已经被绑定了。这种做法与传统音乐产业的潜规则如出一辙——唱片公司趁年轻音乐人懵懂无知时用苛刻合约将其死死套牢。虽然硅谷有 Y Combinator 和 Paul Graham 推广的 SAFE 协议（未来股权简单协议）提供了相对公平的条款保护，但这种诱导年轻人过早做出人生重注的资本行为，其掠夺性本质依然值得警惕。 - 软件工程名言警示：在程序设计中，过早去优化一段尚未定型或无足轻重的代码是典型错误；同理，18 岁为不成熟的点子背负 VC 期望也是人生的过早优化。 - 风险投资的前置入侵：硅谷投资人开出的支票额度从早期的 5 万美元跃升至如今对名校在校生的 100 万美元“预想法”（pre-idea）投资。 - 创业与知识分子的双重特质：主持人 Erik 指出 Amjad 本人罕见地兼具了黑客、动手修补者（tinkerer）的务实行动力，以及广泛涉猎历史、经济、政治与网络迷因（memes）的知识分子好奇心，绝非单一维度的职业工具人（automaton）。 ### [20:19](https://www.youtube.com/watch?v=Q8yFrGwIftE&t=1219s) 好奇心与构建的往复循环：从下国际象棋到“自动化机器学习研究” 面对年轻人如何在务实的职业发展与看似随意的智力探索之间保持平衡的提问，阿姆贾德·马萨德（Amjad Masad）分享了自己思维运转的底层机制。他形容自己总是在两种状态之间往复摆动：一种是“极度渴望亲手做点东西出来”的冲动，另一种则是“疯狂阅读、四处探索新鲜事物”的求知欲。这两者绝非割裂，而是一个相互滋养的递归闭环——在广泛探索中积累了足够的认知与灵感后，他便会迫不及待地想要通过构建产品将其转化为现实。 这种探索过程往往源于毫无功利心的偶然契机。阿姆贾德提到自己近来突然对国际象棋着了迷，这在通常六岁就该入门的棋类运动中显得极为反常（他的六岁孩子最近也开始学棋，而且“眼看就要把我下赢了”）。起初他只是单纯好奇，随即开始写程序解棋谱，继而尝试训练模型。在这个过程中，他发现了一个诡异的现象：即便当前大语言模型在 ARC-AGI 等基准测试上的跑分越来越高、仿佛离 AGI 仅一步之遥，但像 ChatGPT 这样的顶尖模型在下象棋时依然糟糕透顶，频繁出现臆造棋步的幻觉。为了搞清楚原因，从未受过系统模型训练的阿姆贾德一头扎进了机器学习领域，亲自训练模型下棋。 这一钻研让他敏锐地捕捉到了技术浪潮的下一处拐点。正如 2024 年 Replit Agent 问世、在“氛围编码”（vibe coding）概念诞生前夕就让智能体彻底颠覆了编程体验一样，如今的 AI 已经具备了自主执行机器学习研究的能力，即人们常说的“递归自我改进”（RSI）。安德烈·卡帕西（Andrej Karpathy）将其概括为“自动研究”（auto research）或“氛围研究”（vibe research）。阿姆贾德顺着最初对国际象棋的业余好奇，一路推进到了 Replit 的新业务线——让任何业务人员或拥有数据的企业主都能像机器学习研究员一样轻松微调模型以解决实际问题。阿姆贾德感慨道，如果从一开始就要求理性推导，没有任何人能在下第一盘棋时预见到它会演化成一家公司的商业产品线。同样，他沉浸于互联网亚文化与 Meme 的“枝节任务”（side quests），也屡屡为他看清社会趋势与市场盲区提供了独家“阿尔法”（超额认知）。 ```mermaid flowchart TD accTitle: 好奇心驱动的递归创新链条 accDescr: 展示从纯粹的智力好奇心最终转化并孵化为商业产品的递进流程 subgraph Exploration ["纯粹好奇阶段"] A["个人兴趣触点<br/>（中途入坑国际象棋）"] --> B["动手验证与发现问题<br/>（LLM下棋差、易幻觉）"] end subgraph DeepDive ["深入研究阶段"] B --> C["跨界摸索模型训练<br/>（自学训练下棋模型）"] C --> D["捕捉底层范式转移<br/>（AI具备自主做ML的能力/RSI）"] end subgraph Commercialization ["产品与商业转化"] D --> E["业务灵感萌生<br/>（类比 Karpathy 的'自动研究'）"] E --> F["Replit 新产品线<br/>（人人皆可成为ML研究员）"] end ``` ### [24:02](https://www.youtube.com/watch?v=Q8yFrGwIftE&t=1442s) 组织架构的演进与“自动驾驶公司”的雏形 话题转向年轻人如何保持这种探索冲动时，讨论者们指出，多数学生在漫长的中学和大学生涯中，探索的“发痒感”早早被条条框框磨灭殆尽，全身心学会的只有“如何按部就班地顺从”。这种完全服从指令的惯性使得他们很难培养出领导力与自主判断力。针对组织形式的僵化，阿姆贾德提出，随着人工智能的深度介入，未来企业的运作范式将发生根本性变革。 硅谷历史上曾有过多次颠覆传统企业管理结构的激进实验，例如 Medium 推行过的“合弄制”（Holacracy），或是 GitHub 早期尝试的“去管理层化”。这些实验最终大多以极其惨烈的方式宣告失败。然而阿姆贾德认为，过往的失败并非理念本身错误，而是受限于当下的技术水平。在智能体时代，AI 完全可以充当企业内部的“组织黏合剂”，将繁琐死板的行政官僚流程无形地转移至后台自动运转。员工将获得前所未有的真正自主权，得以完全投身于自己最感兴趣、最具创造力的事务中，而其工作成果和实际影响力则由 AI 客观判定。阿姆贾德透露，他正在 Replit 内部积极重塑团队的运作方式，全力探索“自动驾驶公司”（self-driving company）这一愿景，让整个公司最终迈向自我运转。 加根·比亚尼（Gagan Biyani）对此深表共鸣。他在筹办安德森·霍洛维茨学院（Horowitz and Andreessen Academy）的过程中，直接在 Replit 上搭建了整套融资商业计划书，并高频依赖 Claude 与 ChatGPT 加速日常工作流程。加根直言，尽管未来组织演变的确切形态尚无人能精准预言，但可以百分之百确信的是，变革正以超乎常人想象的速度席卷而来。如果年轻一代仍按部就班地沿用工业时代的旧模式去思考和协作，将在未来的商业竞争中陷入天然的劣势。 ### [26:30](https://www.youtube.com/watch?v=Q8yFrGwIftE&t=1590s) 异见者的生存之道：敬畏前辈，但绝不妥协地颠覆过去 在肯定“保持批判与叛逆”的同时，加根进一步厘清了“学会顺从”的真正内涵：一个人需要具备足够的“不合群特质”（disagreeableness），才会有特立独行的见解、敢于成为逆向投资者并不盲从常规；但若过于尖锐偏执，就会演变成无法与任何人共事的破坏性力量。因此，所谓的“顺从”本质上是掌握与他人协作、在现有复杂系统内斡旋并达成最优结果的现实能力。 阿姆贾德从自身经历出发，阐述了“敬畏前辈”与“颠覆规则”之间的精妙平衡。他认为，进入一家公司或行业，应当由衷敬重前人所建立的技术基石，但绝不能迷信他们已经彻底解决了所有难题；相反，新一代人必须敏锐意识到现状的种种缺陷，并拥有大刀阔斧去改变的野心。为了在公司文化中植入这种张力，Replit 将所有的会议室都以计算机与科技史上的泰斗命名： - **马克·安德森（Marc Andreessen）**：网景浏览器（Netscape）的发明者 - **林纳斯·托瓦兹（Linus Torvalds）**：Linux 之父 - **罗伯特·诺伊斯（Robert Noyce）**：仙童半导体与英特尔联合创始人、“硅谷市长” 阿姆贾德向团队传达的核心理念是：“我们敬畏历史，但我们也将毫不妥协、理直气壮地改写历史，即便这会招致外界的愤怒。”他举例谈到，就在 2025 年，他曾因公开发表“人们不再需要学习写代码”的言论而遭遇全网声讨与抵制；然而短短一年之后，当初激烈抨击他的同一批人如今却在泰然自若地表示：“我现在根本不读代码了，全部直接用 Prompt 生成代码。”加根感慨道，硅谷最可贵的一点就在于它对事实的飞速接纳，哪怕某个超前判断在五年前或十年前被千夫所指，人们也能在现实被验证后迅速翻篇、大步向前。 ### [28:46](https://www.youtube.com/watch?v=Q8yFrGwIftE&t=1726s) 奥克兰地方政治与硅谷的高信任度杠杆 谈及从本职工作之外的“侧线任务”中获取养分，加根分享了自己在加州奥克兰（Oakland）参与地方政治志愿服务的非同寻常经历。表面上看，地方政治与创办科技企业风马牛不相及，但加根发现两者的核心技能在双向迁移中极具启发性，而最深刻的一课莫过于认识到了“社会整体信任基础”对协作效率产生的决定性影响。 在非营利机构与地方政治体系中，人与人之间的协作由于缺乏深层信任而显得举步维艰。加根以政治募捐为例剖析了两者之间的残酷落差： - 在奥克兰的地方非营利机构或政治筹款场景中，如果一位出资人在面谈后口头承诺捐助 5 万美元，这笔款项最终能够真正打进账户的概率通常只有可怜的 **20% 到 50%**，口头承诺轻飘飘地缺乏任何实际约束力。 - 相比之下，在硅谷的科技创业圈与投资生态中，类似口头承诺最终兑现到位的概率往往高达 **75% 到 100%**。 这种底层信任度的鸿沟，在根本上决定了一个系统内部的交易摩擦与推进速率。在低信任环境下，组织不得不把庞大的精力耗费在反复求证、相互防备与官僚程序之中；而在硅谷的高信任语境下，创业者与合作者得以凭借口头默契闪电推进，从而爆发出截然不同的敏捷度与生产力。 ### [30:18](https://www.youtube.com/watch?v=Q8yFrGwIftE&t=1818s) 硅谷的高信任文化与放手赋能学生 商业运作中，资金信任能为创业者省去巨大的内耗：若创业者确信资金即将到位，便能在支票实际送达前的六周内全速推进，让后续资金发挥出远高于六周前的价值；反之，若缺乏信任，创始人就会终日焦虑赞助人或投资人是否满意，甚至为了迎合外部压力而偏离原本的使命。这种硅谷特有的高信任机制完全可以移植到教育领域。 硅谷最独特的优势在于，只要有人带着特立独行的想法或颠覆性的做事方式出现，这个生态通常愿意接纳并放手让他们去尝试。尽管这里依然存在基本的社交底线（不能随意撒谎或恶意待人），但如果教育机构能真正信任年轻学生，哪怕大人觉得他们的探索荒诞不羁，也全力予以赋能，年轻人就必定能创造出意想不到的未来，甚至摸索出真正行之有效的下一代组织治理范式。 ### [31:51](https://www.youtube.com/watch?v=Q8yFrGwIftE&t=1911s) 边界探索者：从 Facemash 到漏洞挖掘的引导机制 当谈及如何处理学校里的“捣蛋鬼”（troublemakers）以及打破常规甚至触碰边缘的行为（例如马克·扎克伯格早期在哈佛极具争议的 Facemash，或是黑客入侵行为），学院设立了清晰的底线：严禁违反性骚扰政策或公开霸凌他人，核心在于营造互信与“正和博弈”（positive sum）的环境。同时，安德森·霍洛维茨学院（Horowitz and Andreessen Academy）通过取消打分与考试制度，从根源上消解了学生为了对抗权威而叛逆的阻力。对于类似 Facemash 的项目，学校的态度非常包容——只要不逾越法律红线，即使外界或导师觉得点子古怪，也会全力捍卫学生探索的权利。 这种探索边界的冲动应当被疏导进健康的轨道，而非一味惩罚。最近有新闻报道一名在 Roblox 上自学编程与黑客技术的少年因黑入政府系统被判入狱两年，这引发了关于人才培养的深刻反思。在同 Roblox 创始人兼 CEO 戴夫·巴祖基（David Baszucki）交流时，大家一致认为应当为这类少年建立合法的“沙盒”测试环境，发掘他们的天赋并输送到急需人才的网络安全岗位或美国国家安全局（NSA），而不是直接送进监狱。 ```mermaid flowchart TD accTitle: 青年黑客天赋的正向疏导路径 accDescr: 展示将具有黑客倾向的年轻人从边缘违规行为引导至正向网络安全建设的闭环流程。 subgraph Origin["探索冲动源头"] A["自发黑客兴趣与边界测试<br/>(如 Roblox 编程平台)"] end subgraph Traditional["传统惩戒路径 (负和)"] B["越界入侵政府/商业系统"] --> C["定罪入狱 2 年<br/>人才损耗与对立"] end subgraph PositivePath["教育与产业赋能路径 (正和)"] D["安全沙盒 / 夺旗赛<br/>(如 HackThisSite)"] E["高校与企业漏洞悬赏计划<br/>(Bug Bounty)"] F["输送至关键安全岗位<br/>(如 NSA、企业安全部门)"] end A --> B A --> D D --> E E --> F ``` - **替代性平台与悬赏机制**：过去像 HackThisSite 这样故意预留漏洞供人攻防的网站，其本质如同网络安全领域的“夺旗赛”（Capture The Flag），学校完全可以设立校内安全赏金机制，或鼓励学生参与各大企业的正规 Bug Bounty 项目。 - **教育机制去对抗化**：取消成绩和应试指标后，学生不再需要把精力消耗在与刻板制度的对抗上，创造力与探索冲动便能自然沉淀到实际建设中。 ### [35:40](https://www.youtube.com/watch?v=Q8yFrGwIftE&t=2140s) 培养判断力：何时听从长辈与“非对称收益”权衡 面对“何时听从长辈经验，何时彻底颠覆旧规”的核心命题，最关键的能力在于“判断力”（discernment）。无论是具体规则还是经验法则，其有效性完全取决于具体语境，而判断力唯有通过大量的真实演练（reps）以及承担试错后果才能习得。与之前提出的“技术主义”（technicism）相对，对话中戏称掌握这种平衡艺术为“元优化”（meta maxing）——懂得何时顺应直觉狂飙突进，何时保持敬畏规避灭顶之灾。 这种能力本质上是对“非对称收益”（asymmetric upside）与“非对称下行风险”（asymmetric downside）的权衡。项目制学习（Project-based learning）以及将时间自主权完全归还给学生，核心目的正是提供密集的决策机会，让他们在一次次选择、成功、失败与复盘中，亲身摸索出行动的边界与分寸。 ### [36:38](https://www.youtube.com/watch?v=Q8yFrGwIftE&t=2198s) 硅谷不仅造就技能：道德修养与人际交往的建制化责任 针对外界常年对替代性教育（如 Thiel Fellowship）的质疑——认为传统大学是学生学会“如何做人”、《死亡诗社》（Dead Poets Society）式品格塑造的唯一场所，实际观察给出了有力反驳。在走访过全球各地的难民营与富豪宅邸后，硅谷其实聚集了许多极其友善、深刻且脚踏实地的人。尽管这里也不乏赚了大钱的投机者，但湾区主流的“正和博弈、高信任生态”切实孕育出了极高素养的个体。 人际交往能力是学院设立的绝对支柱之一，两年的学习旨在让学员在心智与为人处世上全面成熟。在旧金山，拥有“靠谱好人”的声誉具有极其真切的商业与社会价值，决定了他人是否愿意托付资本与企业治理权力。 - **告别“自闭天才”人设**：过去硅谷一度盛行并推崇某种“自闭天才”（autistic genius）的人设，例如红杉资本（Sequoia Capital）早年撰写后来又主动删除的 SBF（Sam Bankman-Fried）神话博客，如今这种包装策略已彻底失效。 - **承担建制与公共沟通责任**：硅谷已不再是游离于体制外的反叛群体，它自身已成为社会建制的一部分，必须直接与更广泛的公众对话。正如伯尼·桑德斯（Bernie Sanders）等政界人士也会阅读 Dwarkesh Patel 等科技播客创作者的博客一样，科技从业者在对外发声时必须注意措辞，对缺乏行业背景的公众负起清晰沟通与引导的社会责任。 ### [40:29](https://www.youtube.com/watch?v=Q8yFrGwIftE&t=2429s) 培养适应社会的成年人：从圈内黑话到国会听证的沟通能力 对话由对科技圈封闭表达的批评延伸开来。主持人指出，某些偏激或充满黑话的技术博文并没有真正加深大众的理解，反而容易让不了解背景的外界得出错误结论。因此，像 Horowitz and Andreessen Academy（HAA）这样的新型教育机构，其核心目标之一应当是培养“心智成熟、能够良好适应社会的成年人”（well-adjusted adults）。优秀的青年领袖不应仅仅精通科技圈同温层的“内部黑话”（in-group language），更需要具备走向更广泛的公共领域、甚至能够站在美国国会听证席上与决策者顺畅对话的沟通与表达能力。 --- ### [41:06](https://www.youtube.com/watch?v=Q8yFrGwIftE&t=2466s) 道德罗盘的构建：从功利主义到亚里士多德美德伦理学 在谈及人文教育与价值观塑造时，Amjad Masad 分享了自己独立探索伦理学与道德哲学的个人经历。他曾广泛阅读功利主义（utilitarianism）的经典著作，但最终选择否定了这种纯粹以结果和功用衡量一切的世界观。相反，他通过博览文史经典与深刻自省，建立起自己的道德指南针，发现自己最契合的是亚里士多德的“美德伦理学”（virtue ethics）。其核心信念在于：行动的标准应当是让自己感到自豪，“无论是否有人在注视，也无论最终带来的即时功利影响为何”。奇妙的是，当你坚守美德与行事准则时，其净效应往往反而是实现了全社会的功用最大化。 ```mermaid flowchart TD accTitle: 伦理框架与时代适应性比较 accDescr: 对比亚里士多德美德伦理学与功利后果主义在奇点时代的应用局限与优势 subgraph MoralFramework ["道德哲学取向"] A["功利主义 / 后果主义"] -->|缺陷| B["依赖对未来的精确预测<br/>技术奇点下未来完全不可测"] C["美德伦理学 (Aristotle)"] -->|优势| D["以品德与自省为基石<br/>做令自己骄傲的事而不受外在评价绑架"] end subgraph DynamicWorld ["快速剧变的世界"] E["刚性教条原则"] -->|缺陷| F["世界剧烈演进，僵化教条注定失效"] G["最小化公理化底线 + 持续演进"] -->|主张| H["坚守极少数底层真实准则<br/>随时代前进而调整适应"] end D --> G ``` 面对主持人和嘉宾关于“在剧变世界中 rigid 原则难以维系”的看法，Amjad 进一步剖析了现代道德困境的双重悖论： - **后果主义在奇点时代的失效**：后果主义（consequentialism）完全建立在“预测未来结果”的基础上，但在人类加速迈入技术奇点（Singularity）的当下，未来具有不可预测性，因此仅靠权衡后果根本无法指导当下的行为。 - **刚性教条与演进原则的平衡**：面对风云变幻的现实，死守僵化刻板的哲学体系同样行不通。正确的做法是拥有尽可能小、近乎公理化的“底层真实准则”（ground truths），并随着对世界演进方向的理解而保持原则的动态演变。 - **博雅教育（Liberal Arts）的不可替代性**：阅读伟大著作并思考人生立足点，并非脱离实际的学术玄谈，而是帮助青年理解自我、明晰立身处世之道并展现人格风貌的基石。 --- ### [44:02](https://www.youtube.com/watch?v=Q8yFrGwIftE&t=2642s) 青年才俊的核心特质：深度探究的“极客执念”与人际谦逊 作为 Replit 的掌舵人与年轻创业者的投资人，Amjad 谈到了他在 16 至 20 岁年轻人才身上最看重的特质。首先是拥有极其深入的钻研能力——能够在某一个细分领域扎得极深，以至于“能倒过来教会我新知识”。绝大多数普通人容易心生倦怠、浅尝辄止，或循规蹈矩地走传统道路；而真正打动他的年轻人，能够为好奇心所驱使，沿着一根线索一探到底，即使短期内没有任何实质奖励，也能展现出罕见的坚持不懈（persistence）与强大的内在驱动力（intrinsic motivation）。 与这种专业“穿透力”同等重要的，是健康的待人处事方式。Amjad 观察到，年轻聪明的群体极易陷入“认为全世界所有人都是蠢货”的自大陷阱，甚至面对资深前辈时也表现得仿佛对方一无所知。他强调，持有一针见血的逆向非共识观点固然重要，但必须兼具良好的人际交往能力与基本的谦逊尊重，否则无人愿意与之共事。主持人也总结指出，最理想的青年人才是既敢于在特定领域“尖锐突出”（spike）、热烈地对生僻有趣的主题“深度极客”（nerd out），又能展现真诚、不虚伪的讨喜性格（likable），唯有这样真正品行端正、内心纯粹的创造者，才能赢得世界的长期支持。 ## Notable quotes > [00:00](https://www.youtube.com/watch?v=Q8yFrGwIftE&t=0s) 我们已经摸清了一切运转法则，你只需按部就班学习规矩，成为合格公民，然后找份工作、结婚买房买车过完一生 — Amjad Masad > [00:00](https://www.youtube.com/watch?v=Q8yFrGwIftE&t=0s) 这道围墙之外是什么？群星之间究竟隐藏着什么？ — Amjad Masad > [14:46](https://www.youtube.com/watch?v=Q8yFrGwIftE&t=886s) 过早优化是万恶之源 — Amjad Masad > [20:19](https://www.youtube.com/watch?v=Q8yFrGwIftE&t=1219s) 眼看就要把我下赢了 — Amjad Masad > [26:30](https://www.youtube.com/watch?v=Q8yFrGwIftE&t=1590s) 我们敬畏历史，但我们也将毫不妥协、理直气壮地改写历史，即便这会招致外界的愤怒 — Amjad Masad > [26:30](https://www.youtube.com/watch?v=Q8yFrGwIftE&t=1590s) 我现在根本不读代码了，全部直接用 Prompt 生成代码 — Amjad Masad > [41:06](https://www.youtube.com/watch?v=Q8yFrGwIftE&t=2466s) 无论是否有人在注视，也无论最终带来的即时功利影响为何 — Amjad Masad > [44:02](https://www.youtube.com/watch?v=Q8yFrGwIftE&t=2642s) 能倒过来教会我新知识 — Amjad Masad > [44:02](https://www.youtube.com/watch?v=Q8yFrGwIftE&t=2642s) 认为全世界所有人都是蠢货 — Amjad Masad ## Mentioned Amjad Masad, Gagan Biyani, Erik Torenberg, Replit, Horowitz and Andreessen Academy (HAA), Arthur C. Clarke, 《城市与群星》（The City and the Stars）, 《地堡》（Silo）, 《辐射》（Fallout）, Steve Jobs, Reed College, Macintosh, UC Berkeley, Alpha School, Joe Lonsdale, Joe Lemon, Y Combinator, Paul Graham, Andrej Karpathy, Replit Agent, Medium, GitHub, Marc Andreessen, Linus Torvalds, Robert Noyce, Oakland, Facemash, Mark Zuckerberg, Roblox, David Baszucki, HackThisSite, NSA, Thiel Fellowship, 《死亡诗社》（Dead Poets Society）, Sam Bankman-Fried (SBF), Sequoia Capital, Bernie Sanders, Dwarkesh Patel, Aristotle ## Video description Erik Torenberg sits down with Replit founder and CEO Amjad Masad and Horowitz and Andreessen Academy co-founder and CEO Gagan Biyani to ask what education should look like for a generation growing up with AI. Amjad argues that one of the most valuable things young people bring to society is their willingness to question deeply held assumptions. They discuss how education could create more room for that instinct through project-based learning, intellectual side quests, and giving students the freedom to follow an idea deeply rather than optimizing around grades and credentials. They also explore whether young founders are being pushed to professionalize too early, why Amjad thinks starting a company can sometimes be a form of “premature optimization,” and how curiosity led him from learning chess to experimenting with AI that can conduct machine-learning research. Finally, they discuss trust, judgment, and what it means to develop as a person, not just a builder, including why being contrarian and ambitious still needs to be balanced with the ability to work with other people. Timestamps: 00:00 - Intro 01:04 - "The City and the Stars": Why Young People Should Be Heretics 05:36 - Designing Your Ideal 18-to-22 Experience in 2026 11:20 - Why Curiosity-Driven Learning Actually Works 14:46 - The Counter-Critique: Are Kids Professionalizing Too Soon? 19:40 - The Entrepreneur & the Intellectual: Balancing Both 24:02 - Learning to Follow vs Learning to Lead 35:40 - Discernment & Meta Retard Maxing: When to Trust Elders 36:38 - Beyond Skills: How Do You Actually Become a Good Person? 44:02 - What Amjad Looks for in Young Talent Resources: Follow Amjad Masad on X: https://x.com/amasad Follow Gagan Biyani on X: https://x.com/gaganbiyani Learn more about HAA: https://www.a16z.news/p/introducing-the-horowitz-andreessen Stay Updated: If you enjoyed this episode, be sure to like, subscribe, and share with your friends! Find a16z on X: https://twitter.com/a16z Find a16z on LinkedIn: https://www.linkedin.com/company/a16z Listen to the a16z Show on Spotify: https://open.spotify.com/show/5bC65RDvs3oxnLyqqvkUYX Listen to the a16z Show on Apple Podcasts: https://podcasts.apple.com/us/podcast/a16z-podcast/id842818711 Follow our host: https://x.com/eriktorenberg Please note that the content here is for informational purposes only; should NOT be taken as legal, business, tax, or investment advice or be used to evaluate any investment or security; and is not directed at any investors or potential investors in any a16z fund. a16z and its affiliates may maintain investments in the companies discussed. For more details please see http://a16z.com/disclosures. [Source](https://www.youtube.com/watch?v=Q8yFrGwIftE)

[0:00] Amjad Masad: 我们已经摸清了一切运转法则，你只需按部就班学习规矩，成为合格公民，然后找份工作、结婚买房买车过完一生 [0:00] Amjad Masad: 这道围墙之外是什么？群星之间究竟隐藏着什么？ [14:46] Amjad Masad: 过早优化是万恶之源 [20:19] Amjad Masad: 眼看就要把我下赢了 [26:30] Amjad Masad: 我们敬畏历史，但我们也将毫不妥协、理直气壮地改写历史，即便这会招致外界的愤怒 [26:30] Amjad Masad: 我现在根本不读代码了，全部直接用 Prompt 生成代码 [41:06] Amjad Masad: 无论是否有人在注视，也无论最终带来的即时功利影响为何 [44:02] Amjad Masad: 能倒过来教会我新知识 [44:02] Amjad Masad: 认为全世界所有人都是蠢货

[14:46] Gagan Biyani: The pressure of making a 10-year commitment at 18 isn't for most people. And then the structure of college isn't for everybody either. And so for me, this middle ground of having two years where you can really explore and you can do whatever you want without the pressure... That's super exciting to me. [26:30] Amjad Masad: We respect the past, but we're unapologetically going to change it. [17:17] Donald Knuth: premature optimization is the root of all evil. [1:04] Amjad Masad: zen mindset [17:17] Amjad Masad: premature optimization [20:19] Amjad Masad: vibe coding [20:19] Andrej Karpathy: auto-research [24:02] Amjad Masad: self-driving company [35:40] Amjad Masad: meta maxing [36:38] Amjad Masad: autistic genius [44:02] Amjad Masad: everyone is an idiot

[0:00] Ultimately, the utility of young people [0:02] is to question deeply held beliefs, is [0:04] to be heretics. And in today's college, [0:07] it actually the opposite. It kind of [0:09] beats any outside thought out of them. I [0:12] just really want to see a world in which [0:14] it's okay if you want to go on a side [0:16] [music] quest and go do something that's [0:18] kind of oddball, but also might end up [0:20] being super amazing if it works out. [0:22] >> The future of companies is going to [0:24] change. [0:24] >> The whole industry has changed. It's a [0:27] complete reset on a technology [0:28] perspective. I'd want to go to a college [0:31] that's not prescriptive about what I [0:34] learn. I'm motivated by what skills can [0:36] it give me in order to achieve what I [0:38] want to build. [0:39] >> One of the core tenants of the academy [0:42] sidesteps this problem entirely. We [0:44] don't care if you graduate or not. We [0:47] don't have grades. We don't have tests. [0:48] The amount of weight to push against [0:52] should be a lot less. When you see a [0:54] 18-year-old that is really inspiring to [0:56] you that you're like, I got to spend [0:57] time with this person. What are they [0:58] doing that is capturing your interest? [1:00] >> I think the first thing is that they [1:04] as we were talking uh I I remember this [1:07] uh sci-fi story that I read once uh it's [1:11] called the city and the stars and it's [1:13] about this future city that's governed [1:16] by AGI. We sort of invented AGI. It's [1:19] now governs all of our lives. It solves [1:21] all all problems, all human problems. Um [1:24] there's no wars, there's uh no disease, [1:27] everyone lives forever and people spend [1:30] their time working on like a [1:32] thousand-year art projects and doing all [1:36] sorts of random things to fill up their [1:38] time. But they they found this problem [1:40] where they accumulate so much knowledge [1:43] and memory it becomes like harder to [1:45] keep it all in their heads. So they [1:47] invented this technology where they go [1:50] into some kind of laboratory [1:52] and they compress their memory sort of [1:56] like context compression that we do with [1:58] AI today. They go through all the [2:00] memories and they compress them and then [2:02] they go through a boot camp to like [2:05] retrain them for for uh for for their [2:08] life. So they invented a new way to be [2:14] born. Uh but it it is not you're not [2:17] fundamentally new person. You're still [2:18] the same person. You're just editing [2:20] your memory. But there's a glitch that [2:23] happens in this system in this machine [2:25] and there's a new person that is [2:26] fundamentally born new. And he [2:29] immediately starts causing problems. Uh [2:32] in the boot camp he starts doing weird [2:34] things, new things and everyone's like [2:36] whoa that's that's crazy. Like you [2:38] shouldn't be doing that. That's not how [2:39] we do things. It becomes very [2:41] disruptive. And then when they go out [2:43] into into this like city, he starts [2:46] asking questions like, "What's out [2:48] there? What's in the stars? What's [2:50] outside our city borders?" They're like, [2:51] "We don't talk about outside our city [2:52] borders. It's [2:53] >> sounds like silo a little bit. [2:55] >> It's probably dangerous out there." [2:56] Yeah. A little bit like silo. A little [2:58] bit like Fallout. [2:59] >> Yeah. Um, and and you know, I don't want [3:03] to get into the spoilers too much, but [3:05] eventually that's what he has so much [3:07] drive to change things and go discover [3:11] the stars. That's why it's called the [3:12] city and the stars. He in in in the [3:15] story unfolds in that way and he finds [3:17] all sorts of interesting things out [3:18] there. And I'm reminded of this because [3:21] as we're talking about what is college [3:23] and what is it what is it for? And I I [3:27] think a a related question is like what [3:31] is the utility of young people in [3:33] society? And I think ultimately the [3:35] utility of young people is to question [3:38] deeply held beliefs is to be heretics. [3:42] And in today's college actually the [3:44] opposite. It kind of beats any outside [3:48] thought out of them, right? It it it it [3:51] tells you you're stupid. We've figured [3:54] everything out. You need to learn our [3:56] way and we need to prepare you for the [3:58] outside world so you can get a job and [4:00] be a perfect citizen and go get, you [4:04] know, get married, have kids, and get a [4:06] house and a car and and live happily [4:08] ever after. That world doesn't exist [4:09] anymore. You can do really well in [4:11] college and you can still be unemployed [4:13] with a huge debt. Um, but not only that, [4:17] I think the the the real civilizational [4:21] missing opportunity is is that is is [4:25] that person I in the city of the SARS. [4:29] It took a glitch to get someone with a [4:32] clear mind to be able to question deeply [4:35] held beliefs. And you know, every one of [4:38] us, we probably have a blind spots in [4:41] society. We we have a lot of blind spots [4:43] in Silicon Valley. There's a lot of [4:45] ideas that don't come out of Silicon [4:46] Valley. Uh because there's a monoculture [4:49] here. There's a certain way of thinking. [4:52] Every culture is somehow a monoculture. [4:54] Even the most radical cultures like [4:55] Silicon Valley. And so what what do you [4:58] need from young people? You need them to [5:01] um to experiment. You need to encourage [5:04] them to do to try impossible things. And [5:08] every now and then they're going to have [5:11] a breakthrough [5:12] that once we thought was impossible and [5:15] now it's now it's suddenly possible [5:18] because the technology changed, [5:19] something changed. Um so college [5:22] ultimately should be about that. Let's [5:25] get these these people with clear minds, [5:28] zen mindsets, first principal mindsets [5:31] and encourage them to go question the [5:34] deeply held beliefs. [5:36] No, I think that's well said. I mean, [5:38] imagine if you were 18 today and you [5:42] were had, you know, were able to design [5:44] your ideal experience in terms of, you [5:47] know, how much you go to class to what [5:48] projects you work on to what kind of [5:50] deep study um you would engage in to [5:52] what you'd want to study. Um what are [5:54] some principles or you know that you [5:56] would add to that in terms of how you [5:58] wish if you could redo you know 18 to 22 [6:02] but in 2026 [6:03] >> Yeah. [6:04] >> how you'd think about it. Yeah. I I um [6:06] I'd want to go to a college that's not [6:11] uh prescriptive about what I learn. [6:14] >> Yeah. [6:14] >> Uh I want to be I I want to be an [6:18] environment where there's a lot of [6:19] different paths like you know there is [6:22] um there's a a class about [6:26] um something totally I haven't even [6:29] thought about it's outside my world. [6:31] Right. We're talking about the story [6:33] with um uh Steve Jobs. One of the most [6:37] inspiring speeches he gave is the [6:38] Stanford commencement speech and he [6:41] talked about connecting things [6:42] backwards. And so the way Steve Jobs [6:44] approached college, I think it's the [6:45] ideal experience of college where he's [6:48] like he went to read college but he [6:51] couldn't afford it so he dropped out but [6:54] kept dropping into different classes. [6:56] One of them was calligraphy. [6:58] >> Yeah. and in the calligraphy uh class he [7:00] didn't know why he was interested in it [7:02] and he just did it and learned it later [7:05] on when they were inventing the Mac and [7:06] inventing user u interfaces they're like [7:09] oh we should have uh proportionally [7:12] spaced fonts we should have certain f [7:14] people should be able to select fonts [7:16] and that was a direct kind of u call [7:19] back to his college days learning about [7:21] fonts and calligraphy and things like [7:23] that and and so this like meandering [7:26] that I think you know modern society [7:31] uh thinks is sort of useless. I think [7:33] it's actually the most useful in college [7:35] like can you take a class that would [7:37] like change the direction of your life [7:39] because you discover something totally [7:42] new and outside of your world. So, so [7:45] that's one I would I would I would like [7:47] to have enough freedom to select the [7:50] kind of things I want to study. Um the [7:53] other thing is if I am engrossed in [7:56] building something give me the space and [7:58] time to do it. One of the things about [8:00] my college that was really hard is um we [8:03] had attendance like you they counted [8:06] attendance and if you if you if you [8:08] didn't attend for like six classes [8:09] you're banned. Um and so uh I think if [8:15] you know I'm I'm going to school I'm [8:17] doing all these different things. I'm [8:19] meeting interesting people and then I [8:21] have an idea and I'm like manically [8:23] working really hard. Give me the space [8:25] to do it. Don't like, you know, expel me [8:28] from my [8:29] >> Yeah. Exactly. For sure. [8:31] >> Exactly. [8:32] >> And finally, the the equality of people. [8:34] >> Yeah. [8:34] >> Like I think if you're around [8:35] conventionally minded people, eventually [8:37] you'll be conventionally minded. [8:39] >> Yeah. [8:39] >> If you're in an environment that [8:41] encourages people to question those [8:43] deeply held beliefs, I think you will be [8:45] able to do that. [8:46] >> Yeah. [8:46] >> Yeah. I my I I I went to UC Berkeley um [8:50] go Bears and I I did enjoy it a lot. Uh [8:53] but one of the things that I really [8:55] enjoyed was that there was actually no [8:59] expectation of attending class for most [9:01] situations. And so I had many semesters [9:04] where I had I had like three jobs in one [9:06] semester. And often the time that I was [9:10] expected to be at work overlapped with [9:12] my class. So I just looked at my midterm [9:16] schedule the seme you know in advance [9:19] and I said okay these are my two [9:22] midterms hey boss can I take those two [9:24] days off and otherwise I was working [9:29] >> but it's like interesting because now I [9:31] look back and I I I did enjoy that but I [9:34] feel oftenimes I look at that experience [9:36] and say well what if the environment was [9:39] supportive of this [9:41] >> like what what would happen if Because [9:43] honestly I think today when I talk to [9:46] students you know it's been 20 years [9:47] plus like when I talk to students today [9:51] they all do this. This is like this used [9:54] to be odd [9:55] >> my behavior and now it's come becoming [9:59] the norm. [10:00] >> And the thing that's weird though is [10:03] that [10:04] >> it's not encouraged yet. It's still not [10:07] allowed. And so for me personally, I [10:09] just really want to see a world in which [10:12] it's okay if you go want to go on a side [10:14] quest and go do something that's kind of [10:16] oddball but also might end up being [10:18] super amazing if it works out. [10:20] >> You should be able to use the college [10:21] resources. You should be able to use the [10:23] computer lab, the compute, the bolab, [10:27] whatever it is. [10:27] >> You shouldn't be going against the [10:28] grain. It should be like actually [10:29] actively encouraging you. [10:30] >> Actually, they should be bringing [10:32] resources to you more if you do it. So, [10:35] like if if you have a crazy idea that's [10:37] taking off, like the college, you [10:39] shouldn't be hiding it from your [10:40] college. They should be coming like, "Oh [10:41] my god, like this is amazing. Let's give [10:44] them more." [10:45] >> Totally. There's this kid, uh, Eric, uh, [10:46] I think he's richest friend who has to [10:48] take like, uh, investor calls from the [10:49] bathroom because he's like hiding in the [10:51] in the school. Yeah. They should be [10:53] [laughter] like, you know, coaching them [10:54] through it or Yeah. enabling it in some [10:56] way. Bringing in demo day or something. [10:58] This kind of thing. A couple comments. [11:00] One is going back to your thing around [11:02] like calligraphy. you know people get [11:04] these interests that that will um you [11:07] know might seem random and and might be [11:09] random but in other ways might tie back [11:10] in other ways and I think there actually [11:12] ways in which um the school can uh sort [11:16] of develop like it's hard to um force [11:19] natural interest and so for example I [11:21] was obsessed with basketball as a kid [11:22] but I was also the kid who would ask my [11:24] math teacher like what's the point of [11:25] this like help explain it to me and if [11:27] she could have explained it to me maybe [11:29] I would have got excited about if she [11:30] would say hey you know you're obsessed [11:31] with basketball you you need to [11:33] understand the salary cap or you [11:34] actually basketball is like very [11:35] analytical now in terms of like people [11:37] are using stats and data to evaluate [11:39] players and so if you don't understand [11:41] this you're not going to be a good [11:43] student of the game. That would have got [11:44] me really excited and that's why I'm so [11:45] excited about Alpha School. Yeah. [11:47] >> Is is the way that they take people's [11:48] natural interests and uh get them to [11:51] learn you things they're going to need [11:52] to know over time out of that. Mhm. [11:55] >> The thing that's so hard is we all know [11:58] like the education sort of academic the [12:02] the industry of education if you will or [12:04] the academic uh cannon of education is [12:07] completely on board with this like they [12:09] totally understand that it would be [12:10] great if every student could have a [12:13] customized curriculum, you know, and Joe [12:16] Joe Lemon of Alpha School talks about [12:17] this all the time. The problem is that [12:20] practically it seems impossible, [12:23] but actually I think now is the moment [12:26] when it's completely reverted. And it's [12:28] not just impossible, it's actually [12:31] impossible to imagine why you wouldn't [12:33] do that. [12:35] >> Specifically, like for me, if I think [12:37] about the academy, right? in the academy [12:42] if you are going to learn coding [12:46] I don't want you just to take a class on [12:48] Python no like go build something that [12:52] you're super excited about in Python [12:55] >> you know use it as a tool to learn the [12:58] language and the great thing about today [13:01] you know with AI is every single student [13:04] can go and do that on their own and they [13:06] don't need a customized syllabus or [13:10] exercise given to them. They just need [13:12] the freedom and confidence to say yes, [13:14] you're allowed to do this. Yes, [13:16] absolutely. We're super excited about [13:18] it. [13:19] >> Yeah. I mean, this um you know, the [13:21] buzzword for it is like project based [13:23] learning. There shouldn't be a buzz word [13:25] for it because that should be the [13:26] standard. The idea that you're learning [13:28] something and you don't know the utility [13:30] of it, I think is a very modern [13:32] invention. It's very weird thing to have [13:35] done in society that you would go into a [13:39] statistic statistics class and you would [13:41] never learn that you could use it for [13:42] basketball or you could use it for some [13:44] of your other interests. [13:46] >> Um and so that there was this inversion [13:49] for a long time of how we learn things [13:52] like you would go and you you would [13:53] learn some of the things and you had no [13:56] idea what is the utility, how you would [14:00] you apply it in the real world. Um, and [14:03] and for someone like me, and I think for [14:05] a lot of people that are more [14:07] pragmatically minded, I I'm not [14:10] motivated at all by by by a pure pursuit [14:13] of knowledge. [14:14] >> Yeah. [14:14] >> I'm motivated by skills. What what [14:17] skills can it give me in order to [14:19] achieve what I want to build and achieve [14:21] my dreams? And every now and then, you [14:24] find a teacher that is able to [14:26] >> Yeah. tell you that that's able to [14:28] motivate your learning by telling you [14:30] what kind of skills you can acquire [14:31] because of it. Totally. Well, it's [14:33] interesting because there is you know [14:35] the critique that we're addressing that [14:36] you know um we're solving for which is [14:38] hey there's there's certain schoolings [14:41] environments where there's very little [14:42] practical skills gained or they don't [14:43] train for the real world afterwards even [14:45] though they claim to. But then there's [14:47] this also this opposite critique and I'm [14:48] curious if you guys how we balance this [14:50] which is hey some people are sort of [14:52] professionalizing way too soon and not [14:55] having childhoods or something or not [14:56] you're not spending time to cultivate um [14:59] intellectually or or as a person otherw [15:01] there's still kind of this sort of uh [15:03] ineffable quality that some people think [15:05] you get at a experience like college [15:07] where you're sort of you know with uh [15:09] meeting new people getting new friends [15:11] uh you know sort of going to party I I [15:13] don't know what people are describing [15:15] exactly you studying the great books, [15:18] etc. What do you think about sort of the [15:19] this the stuff that's gained beyond just [15:22] the technical skills of uh of building a [15:25] company or working at a company? [15:26] >> I I think there's this like huge unfair [15:30] choice that a lot of young people are [15:32] provided today. It's either go and get a [15:35] 4-year go to a four-year school where [15:38] there's a lot of structure around what [15:42] you should be doing. You're going to [15:43] have to attend classes. is you're going [15:44] to have to do homework and tests and [15:46] grades and all of that. And then there's [15:49] another option which is you can go and [15:52] start a startup and someone will give [15:53] you $50,000 tomorrow to start it. And [15:57] the problem that I see is look, I think [16:00] there's lots of people who are still [16:01] going to go to college and love it, you [16:03] know, and I think there are lots of that [16:05] there are young people even at 16, 17 [16:09] who should be starting a startup, but I [16:11] think there's an amazing massive group [16:14] of people who I don't know about you [16:15] guys, but certainly for me, when I was [16:17] 17 or 18, I knew I wanted to enter [16:21] business, but it was so nice to have a [16:24] couple of years to explore and learn and [16:27] like no pressure, you know. And so [16:30] personally, I think the investor track [16:32] should be way underutilized like like [16:35] only a very small number of teenagers [16:38] should be raising money instead. And by [16:41] the way, grants are great, right? Do [16:43] something temporary. Make it clear it's [16:44] temporary. That's fine. But the pressure [16:46] of making your a 10-year commitment at [16:49] 18 isn't for most people. And then the [16:52] the structure of college isn't for [16:55] everybody either. And so for me, this [16:57] middle ground of having two years where [17:00] you can really explore and you can do [17:03] whatever you want without the pressure [17:05] and Silicon Valley will come to you and [17:08] provide you with all the information and [17:10] the at resources that you need and when [17:12] you have a cool idea, it's just like [17:14] more more. That's super exciting to me. [17:17] >> Yeah. I mean um I think you know in in [17:20] programming we have this concept of uh [17:22] premature optimization and the idea like [17:25] I think the saying is premature [17:27] optimization is the root of all evil. [17:29] That's how bad people think it is where [17:31] you where you sort of start optimizing a [17:33] program that is not eventually relevant [17:36] or consequential. Um and I think a lot [17:39] of starting startups is premature [17:41] optimization. And I see in a lot of kids [17:43] now where they're part of a group and a [17:46] lot of their friends are raising a lot [17:48] of money and some of them are becoming [17:49] unicorns and they get it just becomes [17:51] the motivation is envy. Um and really [17:54] negative motivation as opposed to [17:56] positive um and then they pre-commit to [17:59] an idea they're not even interested in [18:00] and they raise money for it and they are [18:02] kind of pre-optimizing for some path [18:05] that um that they haven't really reached [18:08] yet. Right? And so that's why I always [18:11] come back to this idea of curiosity like [18:13] what is actually driving you what is you [18:16] know once you find that idea that you [18:18] want to dedicate your life to or [18:20] dedicate the next 10 years to it's going [18:23] to have such gravity you're not going to [18:25] be able to avoid it you know I mean for [18:27] me you know you know I I tried not to [18:31] start Raplet like actively not to start [18:33] it but it kind of exploded into a [18:36] startup on its own right [18:37] >> and that's that's the feeling you should [18:40] get. It's almost like I I I guess I have [18:43] to do it, right? Because it's it's like [18:45] there's so much momentum. [18:46] >> Yeah. [18:47] >> Behind it. Um and I think there's I [18:50] think a predatory thing that happens [18:51] right now where I've heard that uh VCs [18:54] go to Stanford students and they give [18:56] them like a million dollars, not even [18:58] 50K. 50K is like [laughter] a million [19:01] dollars pre- idea and it's like the [19:03] moment you start your idea, you know, [19:05] you already have a cap table. Yeah, [19:07] >> I think that's sort of like the music [19:09] industry. Um, and like, you know, Kanye [19:12] talked a lot about this where it's like [19:13] how predatory it is where you you find a [19:16] young artist and you kind of lock them [19:18] into a contract and they don't even [19:20] understand that contract. Like luckily I [19:22] think in Silicon Valley we have uh the [19:23] safes. Yeah. Which are you know great [19:26] contribution by by by YC and Paul Graham [19:29] but the uh but it's still somewhat you [19:33] know somewhat predatory because you're [19:35] taking this young person you're [19:37] precommitting them to something they [19:38] shouldn't have committed to it. [19:40] >> Yeah. It's really interesting because [19:42] I'm you're an interesting person to [19:43] bring on this podcast in the academy [19:45] more broadly because you're both an [19:47] entrepreneur and a builder and a hacker [19:49] and a tinkerer and very pragmatic, but [19:52] you're also an intellectual. You also, [19:54] you know, very curious, you know, widely [19:56] read across, you know, history and [19:57] economics and culture and politics and, [19:59] you know, you spend a lot of time on [20:01] memes like, you know, like uh you sort [20:03] of um pursue your interest there. to [20:06] going back to this critique. You're [20:07] neither sort of divorced from the, you [20:09] know, economic real real world, but [20:11] you're neither neither that nor a [20:14] automaton, you know, like there are some [20:16] people who are so career focused that [20:18] really that's just like the only thing [20:19] they think about is how to make more [20:21] money or or how to make this thing [20:22] successful. And so I I think to to your [20:25] point, Gogan, sometimes people are [20:27] funneled into these, you know, either of [20:29] these paths. And I'm curious how you've [20:31] thought about is that something that [20:32] you've always been just naturally or how [20:34] can people both be like very pragmatic [20:37] in their career, you know, as as young [20:38] people and yet also make time or make [20:41] space or make interest for um following [20:44] their intellectual curiosity even if it [20:47] might seem, you know, random or, [20:49] [snorts] you know, alien from that. [20:50] >> Yeah. Um I I think that uh I I sort of [20:53] tend to oscillate between having an itch [20:56] to build something uh and [21:00] having an itch to like read a lot and [21:03] explore a lot and see what's what's out [21:05] there, what's interesting and they feed [21:07] on each other. It's a sort of a [21:08] recursive process where I accumulate [21:11] enough knowledge and ideas from all the [21:13] things I explored that I really want to [21:15] go build something with it now, right? [21:17] And and so um like you know for example [21:20] recently I got I got really into chess [21:22] which is like very late in life to get [21:24] into chess. Like usually you get into [21:26] chess when you're 6 years old. Um and my [21:28] kid actually 6 years old is getting into [21:30] chess pretty soon is going to be beating [21:32] me. [laughter] Um but but I I didn't [21:34] know why I was interested. I got [21:35] interested and then I started writing [21:37] programs to like solve chess and then I [21:39] started training models and I was like [21:42] why LLMs like chatbt actually sucks uh [21:45] at chess. [21:46] >> Oh really? Yeah. And I was like, well, [21:48] you know, we're the ARC AGI score is [21:50] getting up there. You know, these models [21:51] are, you know, almost AGI, but why are [21:53] they bad at chess? They hallucinate [21:55] moves and and all of that. So, I went [21:57] into um training NLM to to play chess. [22:01] And uh I spent a lot of time uh I've [22:04] never trained a model before that. I [22:06] spent a lot of time kind of exploring [22:07] how to train models. Uh and I realized, [22:12] you know, some people are now called RSI [22:13] or cursive improvement. uh so sort of a [22:16] buzzword for AI being able to do ML. I [22:20] realized that AI is actually now capable [22:22] of doing ML. So the moment we were in [22:26] like 2024 when rapid agent came out [22:28] before VIP coding as a term was was was [22:31] coined [22:33] >> suddenly agents were capable of like [22:35] coding right and uh in and coding [22:39] fundamentally changed uh since then. Um, [22:42] I think we're at a similar moment when [22:44] it comes to machine learning. I think [22:45] machine learning we're, you know, you [22:47] could call it vibe research, you could [22:49] call it auto research. I think that's [22:50] Karpathy's term. Uh, but um I spent a [22:54] lot of time in that world and then now [22:57] like I uh I'm I'm working on something [23:00] on Replet to to to bring that to to the [23:02] world like where anyone can become a [23:05] machine learning researcher or like any [23:07] business owner who has data can like [23:09] fine-tune a model to kind of solve a [23:10] certain problem. Um, and so [23:14] when I started playing chess, like what [23:16] are the hops you need to make to [23:17] understand that, oh, oh, that's going to [23:19] lead me to build a new business line for [23:21] for reflet. It's just impossible. But I [23:23] just followed something that I was [23:25] interested in. It kind of resulted in in [23:27] in something invariably it does, at [23:29] least for me. But even memes and [23:32] culture, I you know, I I get interested [23:35] in in some random subculture in the [23:39] internet. Um, and I learned something [23:41] about the world. I learned something [23:43] about society. I learned something about [23:44] markets. I learned something about where [23:46] the world is headed. [23:47] >> And I have alpha. You have an itch and [23:50] you listen to the itch. And sometimes [23:51] the itches build something. They just [23:53] learn something and they a lot of kids [23:54] get the itch beaten out of them because [23:56] they, you know, they learn they over [23:58] time to ignore ignore in favor of, you [24:00] know, going through these hoops. [24:02] >> And it's helpful to learn how to go [24:03] through hoops, but it's also helpful to [24:05] learn how to um listen to it. helpful to [24:09] I mean not not more than people do [24:12] currently but I think generally it's [24:14] like in order to lead it's also helpful [24:15] to know how to follow [24:16] >> but I think people spend their entire [24:18] high school and you know and often [24:20] college just learning how to follow just [24:22] learning how to you know someone says [24:23] something you do it [24:24] >> and I think it be yeah [24:26] >> my belief is that the future of [24:28] companies is going to change um like I [24:31] think uh you know in Silicon Valley [24:34] there's been so many experiments to try [24:35] to change the nature of firms terms of [24:37] how companies work. M [24:39] >> if you remember uh [24:40] >> hocray [24:41] >> hocriscy medium did that [24:44] >> GitHub did the like no managers [24:47] >> they all kind of blew up in like [24:48] spectacular ways [24:50] >> but I but I think it was a technology [24:52] problem like I think now like you can [24:55] have agents kind of function as the glue [24:58] within these organization and like have [25:00] the bureaucracy be this invisible thing [25:02] that's happening by machines in the [25:04] background and then people have true [25:06] autonomy to work on the things that they [25:09] find like most interesting and AI can [25:10] sort of judge your impact. [25:12] >> Um, and I don't have fully formed [25:15] thoughts about this, but I see it in our [25:17] company where [25:18] >> I've changed how I run the company and [25:20] we call this idea of like a self-driving [25:22] company eventually like the company will [25:24] like run itself. [25:25] >> Yeah. Yeah. You know, it's funny because [25:27] like I genuinely don't have an opinion [25:29] about where it's going to go, but I know [25:32] it's going to change, [25:33] >> right? I know that that everything we're [25:35] doing now is just feels so different. [25:39] >> Yeah. [25:39] >> Um and in starting the academy, I've [25:41] been spending all this time, you know, I [25:43] built the pitch deck and replet. Thank [25:44] you. [25:45] >> And I've, you know, been using Claude [25:48] and Chachu, like I'm using all of these [25:50] services as a way to speed up my work [25:54] and imagining how that changes an entire [25:57] organization. [25:59] And I guess the thing that I know, the [26:01] thing I feel very certain about is it's [26:03] all going to change. It's going to [26:05] change a lot faster than you think. And [26:07] if you're not like natively [26:09] understanding of it, it's going to be [26:12] hard. It's going to be hard to make that [26:15] shift. [26:15] >> Yeah. [26:16] >> Yeah. Or or or or if you're learning the [26:18] ways of the past. [26:19] >> Correct. then you're kind of locked into [26:22] a a way of thinking and way of working [26:26] that actually will put you at a [26:28] fundamental disadvantage in the future. [26:29] >> Yeah. Back to that following point. I I [26:31] I do think learning how to follow is [26:33] very important. [26:33] >> Let me let me phrase it in a different [26:34] way which is um you want to be [26:37] disagreeable enough such that you have [26:38] interesting ideas. You're contrarian. [26:40] You're you're not just going through the [26:41] motions. You're following your own [26:42] curiosity. But you we both know all know [26:45] people who are so disagreeable they [26:46] can't work with anybody. Yeah. [26:47] >> That they're just so alienating to [26:48] people. So maybe another way of [26:50] following just following it is more [26:51] broader is you you need to learn how to [26:53] work with people. You need to learn how [26:54] to navigate systems. You need to learn [26:56] to um you know get what you want [26:58] [laughter] or or get what what is the [26:59] optimal outcome. [27:00] >> Yeah. Yeah. Yeah, I I think um there is [27:03] this [27:06] important thing of respecting the elders [27:09] in a way, but but also not um [27:15] not I mean fundamentally truly [27:18] understanding that they haven't figured [27:19] everything out. [27:20] >> Yeah. [27:20] >> Um I I do think respect is very [27:23] important. um like going into a company, [27:27] you should respect what they've already [27:29] built, but you should also feel like [27:32] that a lot of things are inadequate and [27:34] go change as many things as as you can. [27:37] So, at Replet, I um I named [27:39] [clears throat] all the conference rooms [27:41] after the uh the great in our industry [27:44] in computer science and and tech. We [27:46] have a Mark Andre room, obviously. Uh [27:49] the inventor of the browser um you might [27:51] know, [27:51] >> of course, but [laughter] [27:54] We have like Lionus Survald, you know, [27:56] Robert Noise, all all these people. And [27:58] what I try to tell people at the company [28:00] is like we respect the past, but we're [28:03] unapologetically going to change it. And [28:06] people are going to be mad at us. Like I [28:08] don't know if you remember like 2025, I [28:11] was like cancelled for for saying you [28:13] shouldn't learn how to code anymore. [28:15] >> And now I I talked to [laughter] [28:18] >> Wow. Yeah. That was only a year ago. [28:22] this a year ago and now the same people [28:24] that were like win against me, they're [28:25] like, "Oh, I don't read code anymore. I [28:27] just like prompted by code." [28:30] >> Yeah. I mean, it's amazing how many [28:32] things I've been wrong about or right [28:34] about that, you know, 5 10 years early. [28:36] And I think the thing that I love about [28:39] Silicon Valley is people just move [28:41] forward. They're okay with it. They [28:43] accept the takes. You know what I mean? [28:46] >> And that's the other thing that I love [28:48] about the culture of the valley. I think [28:51] people, [28:53] you know, so I spend a little bit of [28:54] time, my free time now, volunteering in [28:57] local politics [28:59] >> in Oakland. [laughter] [29:01] >> Yeah, it it's really interesting. It's a [29:03] side quest for sure, right? It's a side [29:05] quest like why would it be relevant at [29:06] all? [29:07] >> Um and actually I think it's it's helped [29:09] me in so many ways in business. Uh the [29:12] the skills are extremely transferable in [29:14] both directions. Um, but one of the most [29:18] interesting things is appreciating how [29:22] valuable trust is and if you have a [29:26] foundation of trust throughout a [29:28] society, how much it changes the way [29:31] that you can move forward. You know, in [29:34] local politics, you know, as a simple [29:36] example, let's talk about nonprofits. If [29:38] you're running a nonprofit and you raise [29:40] money from someone and you have a [29:42] conversation with them and they say, [29:43] "I'll give you $50,000, [29:46] the percent likelihood that that $50,000 [29:49] will land in your bank account is [29:52] somewhere in the 20 to 50% range after [29:55] >> after they've committed verbally, right? [29:57] So verbal commitments don't have weight [30:01] in Silicon Valley. Uh, [30:05] I think it's probably 75 to 100% [30:09] depending on who you're talking to, who [30:12] you are. [30:12] >> Yeah. [30:13] >> And the difference in how you operate, [30:16] like [30:18] I can confidently continue to move [30:20] forward on my business [30:22] if I know that the money's coming. And [30:24] then by the time you've delivered that [30:26] check, I've made like six weeks of [30:29] progress, right? And now all of a sudden [30:32] your check is more valuable six weeks [30:35] later, [30:35] >> right? [30:36] >> And the reverse is also true. If you do [30:39] not have trust, then all of a sudden [30:42] you're like, well, [30:46] what if this donor is upset about this [30:48] thing, let me do this thing differently, [30:49] blah blah, and you're not fulfilling [30:51] your mission. You're worried. You have [30:53] anxiety and stress about this thing. And [30:56] so I think that is something that [31:00] Silicon Valley offers [31:02] >> that can be applied to school. [31:04] >> In what way? [31:06] >> Trusting students [31:09] to do what they want to do and follow [31:11] the threads they want to do is something [31:13] that I think we can uniquely offer. I [31:16] think that Silicon Valley is a place [31:18] where someone who has a unique idea or a [31:20] different way of doing things can [31:22] generally be accepted and allowed to do [31:23] it. Now there are some social rules like [31:26] you can't just go around being mean to [31:28] everyone or lying or it's it's not that [31:31] you know it's not that permissive. [31:32] >> Yeah. [31:33] >> But if we start trusting young kids and [31:37] stu and and young adults to go and do [31:41] follow what they want and then we [31:43] empower them even if we think what [31:44] they're doing is crazy then they're [31:46] going to build that future you're [31:47] talking about. They're going to figure [31:48] out the next version of holocracy that [31:50] actually works. [31:51] >> Let me ask you a question. might be [31:53] tricky, but um [31:56] what are you going to do with the [31:57] troublemakers? Like invariably there's [32:00] going to be people that are going to be [32:01] like breaking things like you know Zach [32:03] famously built like face mash which was [32:06] like really controversial. Um like I [32:08] hacked into someone might [32:11] [laughter] [32:13] >> Yeah. What do you what do you think of [32:14] that? Um I mean obviously there are [32:16] things that really cross some kind of [32:18] red line on criminality but like [32:20] >> Oh sure. Yeah. Yeah, I mean just to be [32:23] clear like you have to draw a line [32:25] somewhere, right? So there's no question [32:27] like if you do something like violate [32:28] our sexual harassment policy or you know [32:31] you are uh super mean to someone in [32:35] front of every you know like there are [32:36] going to be rules of course like we have [32:38] to create a culture of of trust and [32:40] positive sum [32:41] >> right [32:42] >> at the same time I think that one of the [32:44] core tenants of the academy sidesteps [32:48] this problem entirely. Mhm. [32:50] >> We don't have grades. We don't have [32:52] tests. The sort of amount of weight to [32:57] push against [32:59] should be a lot less. And so therefore, [33:03] a lot of these challenges, these things [33:05] that people do, it's like you want to [33:06] build face smash, go for it. Like I [33:08] mean, who cares, right? Like [33:11] I'm willing to defend your right to do [33:14] that. And I'm also happy if someone [33:17] asked me on a podcast to say, "Yeah, [33:18] that was a weird idea." Like, "Well, I [33:20] don't know what that kid was thinking. I [33:22] didn't love it, but you know what? They [33:24] got to do what they got to do." And [33:25] that's how that's how people are going [33:27] to be. And so I'd say the same thing is [33:30] if you create an environment of trust [33:32] and if you allow that to happen and you [33:35] design it so that they don't have a sort [33:37] of like anti- athoritarianism mindset [33:42] then I think you open up a lot of [33:43] creativity and a lot of opportunity. [33:45] >> Yeah. I also think that um there's an [33:48] opportunity to funnel [33:51] that drive to test the boundaries in a [33:55] in a healthy way. Um, [33:58] I was uh reading the story the other day [34:00] about about a kid that hacked into some [34:02] government thing and is like getting [34:03] like two years in prison and he was on [34:06] um on an interview talking about how he [34:08] learned to hack on um uh on Roblox. And [34:11] apparently there's like like a big trend [34:14] where people learn programming and [34:16] hacking on Roblox. Um I was talking to [34:19] Dave uh the founder CE of Roblox. It was [34:22] like be really cool to have [34:25] uh a some kind of sandbox for these [34:28] people to play in and then discover [34:29] talents and then funnel them into cyber [34:31] security jobs. [34:32] >> Yeah. [34:33] >> Or even the government instead of like [34:34] sending these kids to jail like take [34:37] them to work at the NSA or work at like [34:39] some we need cyber security talent right [34:41] now. Like why are you like putting them [34:43] in jail, you know? [34:44] >> Yeah. [34:45] >> Yeah. And there's a line like you know [34:46] there are kids who are hacking who are [34:48] probably doing things that I I wouldn't [34:50] be comfortable with. But there's also [34:52] kids who are doing it for reasons in [34:54] ways that are honestly fine like and and [34:57] if we can just have a little bit more [34:59] permissiveness. I think we can create a [35:01] lot more opportunity for them [35:02] >> and and repurpose that talent as you [35:04] mentioned. [35:04] >> Yeah. I mean uh one one site I was [35:06] really into growing up called hack the [35:08] site. It was hack the site.org. I don't [35:10] know if it's still around. [35:11] >> It's like a website that has [35:13] vulnerabilities built in. [laughter] [35:15] >> Uh like [35:16] >> it's like capture the flag, right? Yeah. [35:19] >> Um, and so like I mean I I think we're [35:22] kind of digging deep into one one aspect [35:24] of it, but like you you can have a [35:26] bounty program in the RC, right? [35:28] >> For sure. [35:29] >> Absolutely. Or I can encourage the [35:31] students to go you do the bounty [35:33] programs at all the other schools if or [35:36] all the other companies because most [35:37] companies have a bounty program. [35:39] >> Yeah. Absolutely. [35:40] >> The um Yeah, it's interesting. I mean, I [35:42] think going back to your thing around, [35:43] you know, when to trust the elders [35:45] versus when to disrupt. It's kind of [35:48] there's this general uh trend in terms [35:50] of like what's really important to gain [35:51] is discernment because all like you know [35:53] rules and heristics. It's like it really [35:55] matters the context and case by and the [35:57] way to get discernment is by having reps [36:00] um and and being trusting the students [36:01] to sort of fail on their own to make [36:03] decisions to to learn from it. It's you [36:05] coined uh you know technicism on this [36:07] podcast. The thing I'll coin is uh meta [36:09] maxing. [laughter] when to [36:11] back versus when to like when to go on [36:14] pure instinct versus when when to sort [36:16] of like when that will drive you into [36:18] into trouble. And I think it relates to [36:19] like when there's asymmetric upside [36:21] versus asymmetric downside. You kind of [36:23] need reps and and I I think that's a lot [36:26] of what project based learning or just [36:28] giving people more control over their [36:30] time is the idea of uh of just giving [36:33] them more opportunities to choose a [36:35] path, succeed, fail, learn from it. [36:38] >> Yeah. Here's my last question. There [36:39] there's some people who would say [36:40] something like hey you know um you know [36:43] remember when teal launched his [36:44] fellowship and a lot of people were mad [36:45] at they were like college you learn how [36:47] to be a good person [laughter] like it's [36:49] it's something beyond this in uh this [36:51] technical you know skill set but even [36:53] beyond this intellectual you know people [36:54] imagine like Robin Williams dead poet [36:55] society I don't know but there's sort of [36:57] this like cultivate you know we're 18 [36:58] you're becoming a person [37:00] >> and so I'm curious if you could talk a [37:01] little bit about how we see sort of in [37:03] the Bay Area you in this environment [37:05] that is both you know hyper um career [37:07] oriented but also you hyper [37:09] intellectual, but what is the sort of [37:11] cultivation of values or morals or [37:13] ethics or or just like becoming an [37:15] adult? Um, what do what do you what do [37:17] we say about that? [37:19] >> Yeah, it's it's it's actually [37:21] interesting. So, I've traveled all over [37:23] the world. I I've uh I would like to [37:25] think, you know, I've been to refugee [37:27] camps. I've been to, you know, billion [37:29] billionaire's houses. I have met people [37:32] from every I've been to [laughter] [37:34] house. Um, I uh [37:38] I would say that some of the kindest, [37:41] most thoughtful, most grounded people [37:45] I've ever met are here in Silicon [37:46] Valley. And it's not everyone. I mean, [37:49] there are there are some real idiots out [37:51] there who've made a lot of money. Like, [37:53] don't get me wrong, I don't have respect [37:55] for everyone who's made money. That's [37:56] not the truth. But the positive sum, [37:59] high trust ecosystem actually, I think, [38:01] breeds really amazing human beings. [38:03] >> Yeah. Um and interpersonal skills is an [38:09] absolute pillar of the academy. Like we [38:13] want we want our uh you know graduates [38:16] or attendees or fellows or whatever [38:18] we're going to end up calling them. We [38:19] want the people who attend the academy [38:22] to be positive influences on society and [38:26] the world. And I think that that's like [38:28] one of the missions of the academy is to [38:32] create a positive sum world. So I [38:36] actually think we absolutely believe in [38:39] a 2-year, you know, you're going to [38:41] spend two years and you're going to grow [38:42] as a human being. You're going to meet [38:44] other people and we're going to create [38:45] an environment where you can see the [38:47] benefits of the culture that I think [38:50] many people don't pay attention to, but [38:52] is very much here in in in San [38:55] Francisco. I I I think having a [38:58] reputation of being a good person here [39:00] actually is is is very valuable and [39:03] people trust you with their money. [39:04] People trust you with governing your [39:07] company and and and and so on so forth. [39:10] Uh and I think developing that that [39:12] sense in university is uh is very [39:15] important and I think being a [39:16] welladjusted person is is important. Um [39:19] I think that um you know for for a while [39:23] like you know I I I think you know some [39:27] people play the role of being like this [39:30] autistic genius person. Um and maybe at [39:34] some point it paid off. Uh like I [39:38] remember the Sequoia blog post about SPF [39:40] that's like gold. They deleted it. You [39:43] can find it on archive.org. Um [39:47] and I don't think [39:49] I I don't think it works anymore. I [39:51] think Silicon Valley is no longer this [39:55] sort of renegade thing. We're kind of [39:57] it's it's an institution. It's part of [39:59] the establishment anyways. [40:00] >> Um and we also have to interface with [40:03] with the world, right? Um and people [40:06] need to understand that we have a [40:09] responsibility [40:11] uh that we're not just talking amongst [40:12] ourselves. Um, and what we're saying, [40:18] like, you know, Bernie Sanders read [40:20] Dwark's blog post [40:22] >> and I I criticized Maresh for kind of [40:24] the language and and it's like [40:28] people who don't understand the [40:29] technology are they're they're going to [40:31] they're not the that blog post in my [40:34] opinion did not create more [40:36] understanding in them. it actually it [40:38] probably if you don't understand it [40:39] probably could you know you you you [40:41] learn the long wrong lessons uh from [40:44] from that that thing and so um so I I [40:49] think it's important that we create or [40:52] this academy creates uh welladjusted [40:56] adults that not can only speak the [40:58] inroup language which is important [41:00] >> but can you know can eventually be in [41:04] front of Congress. Um, is there anything [41:06] I guess lastly you want to say about [41:08] sort of like cultivation of of of values [41:11] or like the sort of the idea of be you [41:13] know being a good person um like is [41:15] there anything there that you feel [41:18] school or school environment contributes [41:20] to or do you think we spoke about [41:21] >> we mentioned like the great books it'd [41:23] be good to like teach some of these [41:25] things like [41:26] >> I spent a lot of time [41:28] >> studying [41:30] uh ethics and morality on my own. [41:32] >> Yeah. Um, I read, you know, a lot of [41:36] the, for example, a lot of the [41:39] utilitarian kind of corpus of and I [41:42] ended up kind of rejecting that. I don't [41:44] think that's the the the way of the [41:45] world. Um but um I sort of like found my [41:50] moral compass by reading wildly and [41:55] introspecting [41:57] um and truly kind of like coming to [42:02] certain conclusions about how I want to [42:04] act in the world. Um and ultimately like [42:07] the way I would describe myself perhaps [42:09] is like closest to like virtue ethics [42:12] >> uh from Aristotle and then and and the [42:14] idea is that actually don't like I [42:17] should act in a way reg that [42:21] uh that I would be proud of. [42:23] >> Yeah. uh regardless of who's watching or [42:25] who's not watching or regardless of the [42:27] of the impact. And I think in a weird [42:30] way if if you're virtuous [42:34] uh and if you have certain principles uh [42:37] the net effect of that is that you are [42:39] utility maximizing. I think if you have [42:41] the right principles [42:43] >> and you need to also change over time [42:44] and you need to kind of figure out where [42:46] the world is headed and and make sure [42:48] you're fitting in it in the right way. [42:50] But all of this to say is that I think [42:52] there is [42:53] >> the there there is value in um in the [42:57] liberal arts and kind of reading some of [42:59] this stuff and and trying to understand [43:02] how you feel about how you want to act [43:04] and how you want to present yourself. [43:06] One thing that just made me more [43:07] sympathetic to you to to [43:09] argument in terms of it being like [43:11] axiomatic is that the world is changing [43:13] so much um that it's hard to have rigid [43:16] principles in a in a world that is uh [43:19] you know or or or a whole philosophy [43:21] that that is rigid um you know in a [43:24] world that is changing [43:25] >> right um and [43:27] uh weirdly the opposite is true as well [43:30] where you can't be a consequentialist [43:34] orian [43:35] when we're entering the singularity and [43:38] you can't really predict the future [43:39] because that's based on [43:42] >> predicting the future and so [43:44] >> it it is a tough time. I think you need [43:46] to have principles that are evolving [43:49] over time. I think that's uh have some [43:51] like ground principle ground truths that [43:55] you hold um dearly and I and they should [43:58] be as as aatic as small as possible. [44:02] >> Yeah. In closing, we're grateful for [44:05] your support of academy. We're excited [44:06] to have people building on Replet, [44:07] potentially interning Replet, [44:08] potentially being being funded by you. [44:10] And so maybe as as a last question, when [44:12] you see a 18-year-old or 20-year-old or [44:14] 16year-old that is really inspiring to [44:17] you that you're like, "Oh, I got to you [44:18] know, spend time with this person or or, [44:20] you know, be involved with them in some [44:21] capacity, what are they doing that that [44:23] is um, you know, capturing your interest [44:26] or what are you looking for in in young [44:27] talent?" I I think the first thing is [44:29] that they dove really deep in something [44:32] they can teach me something. Yeah. [44:34] >> Um [44:36] and and I I I'm interested in like I I [44:40] think I can be interested in anything. [44:41] And I think if you're able to tell a [44:45] story about [44:46] uh some project you've done and you've [44:49] gone really really deep in it, I think [44:51] that's always impressive because that's [44:54] what most people lack and most people [44:56] get bored easily. Most people, you know, [44:58] do the traditional thing. Uh, but did [45:01] you actually follow a thread all the way [45:03] through? That's always really [45:05] impressive. And that shows all sorts of [45:07] good personality traits. [45:08] >> Yeah. [45:09] >> Persistence, intrinsic motivation. You [45:12] might be doing something uh that is [45:14] really uh you're following your [45:16] curiosity and and maybe doesn't have an [45:18] immediate reward. Um [45:22] and um and [45:27] I I think that uh [45:29] uh [45:31] you know [45:33] like you said I mean being contrarian [45:35] having really important opinions but [45:38] it's it's very easy if you're smart and [45:42] if young and smart to think everyone is [45:44] an idiot. [45:45] >> Yeah. uh including me like you know I'll [45:47] see young people I was like like they'll [45:49] talk to me like I don't understand [45:50] anything but uh I think it's important [45:52] to um respect [45:55] uh uh respect to the you the person that [45:59] you're talking having good [46:02] uh good interpersonal skills I think [46:04] it's important otherwise people don't [46:06] want to be around you I think I think [46:07] that's important [46:08] >> totally yeah I think it's it's both not [46:10] being afraid to spike in something like [46:12] a lot of people just show up because [46:13] they're too you know broad and but not [46:15] really amazing at anything and and just [46:18] follow follow that to its conclusion, [46:19] see how good you can get at it, but and [46:20] while also um having a baseline of sort [46:24] of like um you interpersonal skills or [46:27] just sort of like the right beable. [46:29] >> Yeah. Be be likable and and in a way [46:32] that's not like contrived or fake or you [46:34] know uh just like you know it's it's [46:36] truly authentic because people want to [46:37] support people who are good people. But [46:40] the most endearing thing is you can nerd [46:41] out on something obscure and interesting [46:44] and totally. Yeah. [46:46] >> Awesome job. Thanks for Thank you so [46:48] much. Thanks for having me.