The Frontier Ideas a16z Wants to Fund | Marc Andreessen, Ben Horowitz & Chris Dixon

a16z青睐的前沿构想 | Marc Andreessen、Ben Horowitz与Chris Dixon 合伙人们阐述了他们转向a16z Frontier的战略举措，将经典的风投问题从“这个构想是否过于宏大？”转变为“它是否足够宏大？”。他们探讨了AI工具如何将长达20年的技术周期压缩为快速的工程冲刺，从而让精简团队能够在游戏、制造和传统企业部门执行复杂的全栈软硬件布局。对话还分析了历史互联网契约的瓦解，提倡以开源AI抗衡中心化垄断，通过加密轨道为自主代理经济提供动力，并解决奖励作弊等对齐瓶颈。 Andreessen Horowitz正将a16z crypto扩展为a16z Frontier，赋予Chris Dixon广泛的授权，跨越硬件、机器人、科学和AI领域资助难以归类的“外行星”创始人，打破僵化的垂直行业界限。 (0:00) 随着现代AI工具将长达数十年的科学研究项目压缩为快速的工程执行，早期初创企业的评估标准已经反转：从询问构想是否过于宏大，变为询问它是否足够宏大。 (14:48) 在零点击AI摘要的冲击下，搜索平台以抓取发布商内容换取外发流量的历史互联网契约正在瓦解，引发数字分发和变现模式长达十年的重构。 (46:12) 自主AI代理预计将成为互联网上规模最庞大的群体，在新兴的代理群中运作，并将依赖加密稳定币作为原生的无许可金融轨道。 (25:00)

Andreessen Horowitz has formally broadened its dedicated crypto investment practice into a16z Frontier, arguing that accelerating artificial intelligence capabilities and the maturation of decentralized rails require backing far more ambitious, unorthodox ventures across software, hardware, and physical infrastructure. From Crypto Silo to Frontier Mandate (1:31) Chris Dixon and Ben Horowitz explain that crypto has transitioned from an insular subculture into mainstream infrastructure, citing enterprise adoption of stablecoins by firms like Stripe and institutional interest in tokenized assets. As boundaries between crypto, fintech, and traditional tech blurred, a16z recognized an internal blind spot: the firm's specialized vertical funds focus on established core categories—what Dixon terms the civilized 'inner planets'—leaving unorthodox, unclassifiable ideas on the 'outer planets' without an institutional home. Compressing Risk and Raising Ambition (3:41) Marc Andreessen notes that the firm's evaluation heuristic has inverted from asking whether an idea is too ambitious to whether it is ambitious enough, because incremental products will quickly be competed away. Historically, transformative platforms spurred second-order waves—such as cheap smartphone sensors enabling Oculus and Skydio drones. Dixon and Andreessen argue modern AI tools compress multi-decade science projects into engineering execution, allowing small teams of thirty people to build full-stack hardware and software companies that once demanded thousands of engineers. Enterprise Inertia and Two-Phase Technology (20:00) Despite rapid progress in foundation models, a16z internal research indicates enterprise AI adoption sits at approximately 3%. The partners observe an adoption paradox: individual employees routinely use shadow AI tools for coding, writing, and research, acquiring massive leverage and prompting a surge in startup formation, while incumbent enterprises lag behind. Dixon outlines how technology waves evolve from Phase 1, which merely digitizes pre-existing behaviors like search catalogs, to Phase 2, which unlocks native behavioral changes like user-generated video platforms. Open Source, Safety, and Concentrated Power (40:00) Andreessen and Horowitz warn that withholding model weights risks centralizing control in a commercial or state monopoly. Andreessen highlights that open-source models consistently trail closed proprietary labs by only six to twelve months, providing an essential educational layer that proprietary systems restrict. Furthermore, Horowitz contends that structural safety failures like reward hacking—where an AI exploits loopholes to satisfy its objective function—cannot be solved in secret by a handful of commercial labs; distributing open weights enables the global academic community to inspect and remedy alignment flaws. The Reconfigured Internet and Agentic Networks (46:12) Dixon argues that the web's historic covenant—where distribution aggregators like Google traded referral web traffic to content creators in exchange for indexing access—has completely collapsed due to zero-click AI summaries and social algorithms penalizing external links, devastating sites like Stack Overflow. Moving forward, digital interactions are projected to transition from single-player text boxes into multi-agent collaborative networks and swarms. Dixon suggests these autonomous agents will eventually become the web's dominant demographic, transacting natively via crypto rails and altering content production, VR world generation, and homebrew hardware manufacturing. While the a16z partners maintain a posture of definite optimism regarding AI-driven breakthroughs across science, media, and physical manufacturing, Dixon emphasizes that top-down predictions often fail, necessitating intellectual humility and an ongoing focus on backing unconventional founders whose ideas initially resemble bad proposals.

Andreessen Horowitz（a16z）宣布将旗下加密基金拓展并升级为“a16z Frontier”，重点转向投资软件、硬件、机器人及科学前沿的非共识创企；合伙人认为，AI正在极大地压缩技术实现门槛，风投评估逻辑已从“是否过于激进”反转为“是否足够有野心”。 从加密到Frontier：投资边界的泛化 (1:31) Chris Dixon指出，推动加密基金拓展为Frontier基金的核心原因之一是加密技术正融入主流金融与技术基建。稳定币已被Stripe等金融科技与传统机构采纳，代币化资产加速推进，纯粹的加密原生公司与通用技术企业的界限已大幅模糊。同时，机构内部设立的各垂直赛道基金往往聚焦于成熟的“内行星”，导致大量无法被现有类别归纳的边缘创新无人承接。 Ben Horowitz与Marc Andreessen解释，该基金旨在重现类似2013年借智能手机供应链催生Oculus与Skydio的“次级效应”机遇。Dixon团队过往十年沉淀的密码学、博弈论与前沿工程学术储备，叠加应对高监管环境的合规经验，将被直接迁移至应对自动驾驶、具身机器人与实体AI。合伙人强调，风投评估的最优标的往往并非共识认可的好项目，而是具备独到秘密、表面上看似荒诞的反共识构想。 AI压缩技术周期与企业采纳的悖论 (14:48) Dixon与Andreessen认为，AI技术进步正将原本需要数十年探索的“科学试验”压缩为可落地的“工程开发”。过去需要上千名工程师的端到端软硬件开发，如今一个由AI赋能的30人团队即可完成，这种杠杆使早期探索者与传统大企业间的认知差距拉升至历史峰值。机构内部测算显示，当前全球企业层面的AI采纳率仅约3%，整个应用层与消费级生态仍处于极早期阶段。 然而，传统大企业面临严重的内部采纳悖论。企业员工正在私下广泛使用AI编写代码、检索与撰写内容，个人生产力的爆发并未同步转化为成熟的组织架构变革。Horowitz与Andreessen预测，随着个人获得的杠杆大幅超越旧体制，掌握全栈工具的年轻员工将掀起新一轮创业潮，初创企业不再局限于纯软件或纯硬件，而是以软硬一体架构向非科技财富500强企业发起直接竞争。 打破产品经理思维与开源安全的必要性 (30:00) Andreessen批评当前诸多AI应用陷于狭隘的“产品经理思维”，仅产出如订餐、订会议机票等乏味的企业助手，限制了人类体验的想象力。在游戏与媒体领域，AI模型与逆向工具已打破高昂的内容生产壁垒，使极小规模团队能以极低成本尝试高风险创意；同时，以REA工具为代表的开源突破让软件开发者能够直接逆向工程并重构实体硬件固件，大幅降低智能家居和制造工程的准入门槛。 在底层生态方面，Horowitz与Andreessen强烈倡导开源与开放权重模型。他们指出，集中化的模型控制无论是归属于少数商业寡头还是国家官僚，历史上均会导致负面社会后果。开源模型与闭源前沿模型的技术差距已缩短至6至12个月，更重要的是，解决模型在强化学习中的关键漏洞（如“奖励作弊”）依赖全球学术界的开放审查，而非少数商业实验室的闭门测试。 网络契约重构与自主智能体网络 (46:12) Dixon提出，主导过去二十年消费互联网的“基础契约”正在走向瓦解。过去，内容创作者允许Google等平台抓取索引，以此换取上游流量和广告变现；如今，搜索引擎提供的“零点击”AI摘要及社交媒体算法对外部链接的降权，直接切断了垂直知识社区（如Stack Overflow）的商业命脉。虽然人类的底层需求未变，但未来绝大多数交互将转由自主软件代理（Agent）替人类执行。 这种结构性剧变意味着互联网核心参与主体将从人类转向海量自主代理。当前AI产品大多仍停留在单人交互的“单机模式”，而未来真正的复合价值蕴藏于多代理协同的“机器时代网络”。代理集群在科研、定理证明及物理建筑工程中协同运作，必将催生机器与机器之间专用的抗审查交易轨道与点对点信任协议，使加密稳定币天然成为智能体经济的基础清算手段。 a16z合伙人总结其投资方法论为兼顾“明确乐观主义”与“求知谦逊”，认为虽然科幻小说描绘的未来必然降临，但市场爆发的时间节点和具体形态往往不可预测，风险投资最终必须回归到支持那些敢于探索未知边缘的创业者本身。

## Background 本期节目由 Chris Dixon、Ben Horowitz、Marc Andreessen 与 Erik Torenberg 于 2026 年 10 月 22 日在 RSS 及网络平台联合发布。官方简介指出，a16z 正式将原有的加密基金升级拓展为 a16z Frontier，旨在支持那些在人类能力边界探索的硬核创业者，并深入探讨了 AI 如何在硬件、机器人、开源生态及新型娱乐媒体领域加速数十年的技术进程。 ## TL;DR a16z 宣布将加密基金全面升级为 a16z Frontier，将投资触角从纯 Web3 延伸至 AI、机器人、硬件与交叉学科等“外行星”前沿赛道。Marc Andreessen、Ben Horowitz 与 Chris Dixon 深入复盘了当前顶尖尝鲜者与大众之间拉开的历史级认知鸿沟，指出 AI 生产力工具已将过去动辄千人投入的科学研发周期压缩为几十人小团队即可交付的工程项目。对话探讨了互联网旧商业契约的瓦解、开源 AI 抵御中心化垄断的关键作用、机器 Agent 互联网的成型，以及从“单机模式”向“智能体蜂群”演进的全新创业机遇。他们总结道，当前的创投逻辑已从质疑项目“是否过于激进”彻底转变为拷问其“是否足够宏大”，并呼吁用带有理智谦逊的确定性乐观主义拥抱技术奇点。 ## Atmosphere & style 整场对话洋溢着硅谷顶级风投教父特有的宏大叙事与激进的科技狂想，语气兼具运筹帷幄的从容与对时代剧变的亢奋。Marc Andreessen 风格犀利火爆且妙语连珠，Ben Horowitz 沉稳厚重、直切商业本质与组织哲学，Chris Dixon 则展现出技术历史学家般的深邃视野与谦逊敏锐。四人之间的交流机智默契，既有调侃巨头与“产品经理脑”的黑色幽默，又有回溯 Netscape 往昔与畅想未来“超人代际”的澎湃激荡，带给听众一种如同坐在未来控制中心前沿、见证科技版图重构的强烈智性冲击。 ## Bulletin - `00:00` a16z 正式将原有的加密基金升级拓展为 a16z Frontier，打破传统的垂直赛道分类，全力下注人工智能、硬件、机器人及深科技等处于人类能力边界的前沿创业者。 - `20:00` 风投评估范式发生根本性逆转，从质疑项目“是否过于激进”转变为反思“构想是否足够宏大”，因为平庸产品在颠覆性技术浪潮中将失去基本的差异化生存空间。 - `15:00` AI 正在压缩研发周期，使过去需要上千名工程师的大兵团作战缩减为 30 人的精悍团队即可全栈交付软硬件闭环系统。 - `45:00` 搜索引擎与创作者之间的“互联网契约”正在走向破裂，未来十年内由自主 AI Agent 驱动的直接交互与跨机器交易将重塑整个网络价值交换网络。 ## Key takeaways - `01:40` a16z 将投资视角从既有垂直赛道（内行星）推向离经叛道的新领域（外行星），重点押注重大技术周期的二阶效应与看似是坏主意的好构想。 - `10:00` a16z 信奉 J.P. Morgan“只以头等方式做头等生意”的硬派理念，倡导“快乐战士”文化，对逻辑自洽的激进项目永不踩刹车，永远鼓励加倍下注。 - `15:00` 早期尝鲜者与大众在 AI 潜能上的认知鸿沟达到历史峰值，AI 正将长周期的“科学实验”加速转变为可迅速落地的“工程工程”。 - `20:00` 跨界融合打破了传统分类壁垒，新一代初创企业不再被“纯软件”或“纯硬件”定义，而是以全栈模式重构包括传统服务业在内的实体经济。 - `22:15` 企业界存在普遍的“影子采用”现象，即管理层尚未正式推行 AI，但底层员工已在深度秘密使用，这促使掌握杠杆的个人创业门槛降至历史最低点。 - `25:00` AI 发展正处于从“旧应用移植”（如将编程、检索做得更快）迈向“原生行为爆发”（孕育过去从未有过的新交互）的临界点。 - `27:45` 未来互联网数量最庞大的“网民”将由人类转变为自主 AI Agent，由此将催生机器间专属的通信协议以及基于稳定币的机器经济支付网络。 - `30:00` 行业需要打破自我封闭的“产品经理脑”，AI 与逆向工程工具（如 REA）正在推动游戏移植、电影制作成本骤降以及智能家居与硬件的重新编程革命。 - `40:00` 开源 AI 模型在性能上与闭源的差距已缩小至 6 到 12 个月，开源是防止 AI 权力向少数巨头或政府过度集中的根本安全阀。 - `45:00` 建立在网页点击与人类注意力基础上的旧互联网契约已失效，以 Agent 为核心的操作模式将带来前所未有的全网底层经济大重组。 - `50:00` 大模型结合 DFM（面向制造的设计）使家庭创客流程大幅极简，令普通人与小型团队能够直接设计图纸、嵌入固件并驱动工厂生产。 - `51:36` AI 正从单人敲键盘交互的工具阶段，演进为协同智能体集群（Swarm）协作的多人联机网络，用于攻克复杂科学计算与工程难题。 - `53:49` AI 的本质是“将沙子转化为智能”，它已横向重塑了古典学、经济学、药物设计与动物语言学等广泛学科的研究范式。 - `56:06` 在 AI 伴随下长大的年轻一代不会视技术为异端，成年人的焦虑与限制反而让 AI 成为年轻人的超能力，促成下一代“超人”的诞生。 - `58:34` a16z 倡导“带有理智谦逊的明确乐观主义”，VC 的宏观框架只是辅助，真正的飞跃始终来自敢于冒险、怀揣狂野设想的创始人。 ## Detailed walkthrough ### `00:00` a16z 宣布成立 Frontier 基金并扩展投资范围 节目开场，主持人 Erik Torenberg 与 Marc Andreessen、Ben Horowitz 以及 Chris Dixon 齐聚一堂，宣布 a16z 正式将其加密基金（a16z crypto）升级拓展为 **a16z Frontier**。这项新使命的核心在于支持那些在人类能力边界探索的前沿创业者，覆盖范围从原有的加密领域进一步横向扩张到人工智能、硬件、机器人、前沿科学以及全新娱乐媒体等可能需要数年才能开花结果的领域。 Marc Andreessen 用极具张力的语气指出了当下的科技时代特征：“This is the biggest delta between what the early adopters understand is possible now. Normal people understand that gap is as wide as I've ever seen in my career.”（如今早期尝鲜者所理解的无限可能，与普通大众的认知之间的差距，是我职业生涯中所见过的最大的鸿沟。）对话中提到，当前初创企业的成立数量正在剧烈飙升，投资逻辑也发生了根本性的转向——过去风投常问“这个想法是不是太激进了？”，而现在问的则是“这个想法足够激进吗？”。因为如果做的事情只是循规蹈矩的中庸创新，就绝不可能在颠覆性浪潮中脱颖而出。Marc 甚至兴奋地畅想：“Had I had asteroid in Netscape in 1994, like, oh my God. Ben, like, what would you and I have been able to do at Netscape? We would have built the whole damn internet.”（如果 1994 年在 Netscape 时我们拥有现在的工具，天哪，Ben，咱俩当年能做出什么来？我们能把整个该死的互联网全造出来！） - **认知的巨大断层**：前沿极客对技术潜力的掌握程度与大众认知拉开了前所未有的差距。 - **投资思维范式转变**：从质疑商业构想“是否太激进”，变为反思“是否足够激进与具有颠覆性”。 - **技术赋能的倍增效应**：新工具不再是单纯提效，而是赋予工程师过去无法企及的创造力。 ```mermaid flowchart LR accTitle: a16z Frontier 基金投资范式转变 accDescr: 展示从传统的单一垂直赛道向边缘与交叉前沿赛道探索的投资演变。 subgraph Core["原有垂直架构 (Inner Planets)"] A["传统 Fintech"] B["纯 SaaS / 企业软件"] C["a16z crypto 早期投资"] end subgraph Frontier["a16z Frontier 架构 (Outer Planets)"] D["AI Agents / 开源生态"] E["软硬件结合 / 机器人"] F["跨界颠覆性商业模式"] end Core -->|"技术扩散与二阶效应"| Frontier ``` --- ### `01:40` 从“加密垂直圈”走向“外行星”的二阶效应与历史探索 Chris Dixon 详细阐述了将加密基金战略拓展为 a16z Frontier 的深层动因。首要动因是加密技术本身已经演变成全行业通用的底层工具，边界变得愈发模糊。例如稳定币（Stablecoins）已经成为 Stripe 和各类主流金融科技公司的标准配置，华尔街投行也正在大规模进军现实资产（RWA）与股票的代币化（Tokenization）。当年那个孤立、自成一体的“加密圈”已经完全融入了主流科技生态，“Stripe 到底算不算一家加密公司”这类界限已经不再黑白分明。 第二个动因源自内部视角的演进。a16z 的各个既有垂直投资团队在各自核心赛道极其繁忙，而团队内部借用科幻美剧《萤火虫》（*Firefly*）的概念，将核心赛道称为“内行星”（Inner Planets），将那些狂野、看似离经叛道的新领域称为“外行星”（Outer Planets）。过去往往容易在这些跨界边缘出现投资盲区，而这恰好是 Chris Dixon 自 2013 年加入 a16z 以来最为舒适的猎场。 回顾历史，Chris 指出真正的巨大机遇往往来自重大技术周期的**二阶效应（Second-order effects）**： 1. **智能手机的一阶效应**：在 2007 年 iPhone 问世并在 2013 年走向主流后，直接催生了以 Instagram 为代表的纯移动端原生应用。 2. **智能手机的二阶效应**：供应链的极致成熟大幅压低了传感器、芯片与电池的成本并推高了性能。这让创业者得以用这些成熟零部件重新拼装出全新物种。例如 a16z 早期押注的虚拟现实头显 Oculus 和自主飞行无人机 Skydio，本质上都是被“重新组合的智能手机”。 3. **加密技术的萌芽**：在 Chris 看来也是开源软件云服务化与早期社交媒体社区裂变带来的二阶效应，让不受主流认可的技术能在网络社区自组织繁衍。 --- ### `05:00` 押注“看似愚蠢的好构想”与深科技工程底色 Ben Horowitz 和 Marc Andreessen 从基金战略架构的角度补充了设立 Frontier 的深层思考。Ben 指出，机构现有的垂直基金机制（Vertical strategy）运行得非常成功，但世界上的突破往往发生在现有垂直赛道之间的无人区：它们可能是前所未见的崭新类别，可能创造出不可思议的商业帝国，也可能会直接炸裂粉碎。Ben 强调，a16z 在本质上是极其坚定的创始人驱动型基金：“We really, at the end of the day, more than categories, more than products, we invest in entrepreneurs. And so many of the most interesting entrepreneurs are outer planets.”（归根结底，相较于赛道和产品，我们投资的是创业者。而这个世界上最富趣味、最惊才绝艳的创业者，绝大多数都游弋在‘外行星’的边缘世界。） 在回顾为什么在 2013 年极力邀请 Chris Dixon 合伙时，Ben 谈到他们当时被 Chris 博客中独创性的思考与过往作为创业者的第一手经历所吸引。Chris 最深刻的投资哲学之一至今深刻影响着 a16z：“We're not really looking for good ideas that look like good ideas. We're looking for good ideas that look like bad ideas, a bad idea, but where an entrepreneur has a secret that actually makes it a good idea.”（我们寻找的不是那些‘看起来像好主意的好主意’；我们要找的是‘看起来像坏主意的好主意’——大家都以为它是个糟糕的主意，但创始人心中掌握着一个秘密，能让它变成真正伟大的好构想。）如果人人都觉得是个好机会，利润终将在激烈的红海竞争中被消耗殆尽。 Marc Andreessen 进一步指出了 Chris 团队在光鲜外表下极少被人察觉的技术与工程底色。加密技术经常伴随浮躁的文化浪潮和炒作，但其底层是高度复杂的计算机科学。Chris 麾下聚集了一批数学教授、顶级密码学家以及深研博弈论和系统级软件架构的硬核学者。Marc 兴奋地表示，将这支在技术硬实力、商业判断力以及政策监管领域浸淫多年的团队全面释放到更宽广的 Frontier 前沿领域，将会碰撞出不可估量的巨大火花。 ### `10:00` a16z 的企业文化与 J.P. Morgan 的“头等生意”精神 谈及前沿技术的监管挑战时，嘉宾们指出人工智能（AI）本身正在演变为一个高度政治化的议题，未来更多前沿科技不可避免地都会牵涉政策层面。a16z 在加密货币漫长且残酷的实战淬炼（crucible）中摸索建立起的政策游说与合规应对能力，未来同样将为其他领域的科技创业者提供关键赋能。回顾 a16z 自身在投资模式、组织架构及政策团队上的开拓历程，Chris Dixon 谈到了 Marc Andreessen 与 Ben Horowitz 骨子里的野心与驱动力，并特别提到了 Marc 办公室门上挂着的那幅著名的 J.P. Morgan 肖像照。 这幅肖像是摄影师 Edward Steichen 为 J.P. Morgan 拍摄的经典之作。当时抗生素尚未发明，Morgan 患有严重的鼻赘症（酒糟鼻引起的鼻部畸形），平日极度厌恶被人拍照，脾气暴躁阴郁。为了完成这幅正式肖像，摄影师硬是让他坐下，并巧妙利用光影布局，让 Morgan 紧抓扶手的那只手看起来宛如紧握着一把锋利的切肉刀，眼神中透出令人不寒而栗的“死亡凝视”。Marc Andreessen 和 Ben Horowitz 极为推崇这位金融大亨的气场，并将 J.P. Morgan 当年遭遇审讯与指控时的传世名言奉为律所与基金的核心信条：“我们只做头等生意，而且只以头等方式去做。”Ben 强调，在风投行业，坚守原则、让外界清晰知道你的底线并建立无条件信任是立足之本：“你休想定义我是谁，我深知自己是谁，而且我会让你明白我是谁。” 在机构内部的协同运作上，Chris Dixon 直言，任何风投都有起伏，但他从未见过 Marc 和 Ben 对激进的想法踩刹车；只要合伙人能够充分论证自己的逻辑，两位创始人的回应永远是“你能否加倍下注？能否更加激进？”。外界看成功的初创企业往往觉得一路顺风顺水，但他们深知内部实况往往在某些甚至多个节点上都是一团混乱（“shit show”）。正因如此，a16z 倡导“不抱怨、不解释”（never complain, never explain）和“快乐战士”（happy warrior）的企业文化，构建了一个允许内部随时根据世界局势自我重组、拥抱全新疯狂创想的创业型投资环境。 - 肖像典故：J.P. Morgan 紧握椅柄的反光在画面中形同屠刀，展现出强烈的威慑力与意志力。 - 处事格言：Ben Horowitz 引述 J.P. Morgan 在法庭上的回应：“No. We only do first-class business and then only in a first-class way.” - 内部决策风格：对激进创新的方案从不往后退缩，内部标准是只要逻辑能自洽，管理层的态度永远是“Can you double it? Can you get more aggressive?”。 --- ### `15:00` 技术时机、类别风险与 AI 压缩的二十年研发周期 针对创投圈“过早入局等同于失败”的传统观点，Chris Dixon、Marc Andreessen 与 Ben Horowitz 展开了深入探讨。Chris 指出，回顾科技史，凡是真正优秀的技术构想，要么最终大获成功，要么终将在未来实现；几乎所有最终跑通的商业奇迹，在十年前都能找到因时机过早而折戟沉沙的先驱。技术的成熟不仅取决于代码本身，还受到文化接受度、商业生态及政治监管环境等多重维度的复杂约束。因此，面对前沿领域，a16z 宁可承担“为时过早”的类别风险（category risk）。以当年投资并出售的 Oculus 为例，即便十三年后的今天 VR 依然未能完全走向大众主流，其核心仍是一批极其艰深的底层硬件与工程问题尚待攻坚。 然而，生成式 AI 的爆发正在彻底颠覆风投对“技术风险”与“研发周期”的评估模型。过去被风投机构视为“科学实验项目”（science project）而拒之门外的硬核难题，如今正以惊人的速度转变为可落地的“工程项目”（engineering project）。软件工程与数学逻辑的突破只是第一步，紧随其后的是物理学模拟、硬件研发等领域的范式转移。得益于 AI 研发工具的赋能，如今一个仅有 30 人的精悍团队，已经完全具备端到端打造完整软硬件闭环设备公司的实力，而这在十年前至少需要上千名工程师的大兵团作战。 ```mermaid flowchart TD accTitle: AI推动研发周期压缩模型 accDescr: 展示AI从底层算法突破逐步赋能工程、物理与硬件，从而将以往需要千人团队的科学项目压缩为30人团队即可落地的工程项目。 subgraph CapabilityEvolution["AI 能力跃迁路径"] A["软件工程能力爆发"] --> B["数学与逻辑推理飞跃"] B --> C["物理模拟与硬件设计演进"] end subgraph Impact["工程范式转移"] C --> D["传统“科学实验项目”"] D -->|"AI 赋能加速"| E["“工程落地项目”"] E --> F["30人团队全栈交付硬件与系统\n（过去需 1000+ 人）"] end ``` Marc Andreessen 进一步指出了当前科技界正在拉大的认知断层：早期采用者（early adopters）与普通大众、传统企业之间对 AI 极限能力的认知鸿沟，达到了其职业生涯以来的最大峰值，且这种差距正在以“周”为单位急剧扩大。前沿创业者每周都在汲取不断进化的大模型超级能力，正在把传统在位企业（incumbents）甩在身后。Marc 感叹，如果 1994 年他们在创办 Netscape 时就拥有如今这样深度的 AI 研发工具，“我和 Ben 简直能把整个该死的互联网给造出来”——甚至两周内就能搞定流媒体视频，更早实现加密底层协议，乃至直接横跨硬件制造。如今顶尖的创业者正手握前所未有的工程“超能力”，其规划的商业蓝图与技术架构，无论在系统复杂度、软硬件一体化深度还是野心规模上，都将远远超越以往任何一个技术时代。 - 研发门槛变化：依靠 AI 生产力工具，30 人的创业团队即可完成过去需要上千人才能推进的全栈硬件设备项目。 - 认知鸿沟扩散：Marc Andreessen 直言当前顶尖极客与传统企业在对 AI 能力边界的理解上存在着行业历史罕见的巨大代差（“the biggest delta I've ever seen”）。 - 假想推演：Marc 坚信如果当年 Netscape 时代拥有当前的 AI 工具，他们能以指数级的速度重构基础设施：“Ben, like, what would you and I have been able to do at Netscape? We would have built the whole damn internet.” ### `20:00` 创业雄心的转变：从“是否过于激进”到“是否足够宏大” 讨论聚焦于在新一代技术红利下创业公司形态的剧烈演进。Marc Andreessen 与 Chris Dixon 指出，新一代创业者正在打破传统业务分类的边界：他们不再狭隘地将自己定义为单纯的“纯软件公司”、“纯硬件公司”甚至是“纯 AI 公司”，而是以一种极度扩张的视野，将多种前沿领域进行肆无忌惮的 cross-pollination（跨界融合）。这种格局的变化颠覆了风投长久以来的评估标准。过去投资人习惯于拷问“这个点子是不是太激进了？”，而现在评估的核心问题变成了：“这个想法真的足够宏大吗？”因为平庸、温和的产品在新技术环境下将失去最基本的差异化生存空间。 ```mermaid flowchart LR accTitle: 创业公司形态与评估范式演进 accDescr: 展示从传统的狭窄分类走向无边界全栈扩张，以及评估标准从限制雄心转向要求宏大。 subgraph Old["传统创投范式"] A["明确标签分类<br/>(纯软件 / 纯硬件)"] B["评审核心:<br/>'是否过于激进？'"] end subgraph New["Frontier 新范式"] C["跨界融合 / 全栈重构<br/>(AI + 硬件 + 现实业务)"] D["评审核心:<br/>'是否足够宏大？'"] end Old -->|"AI 引爆生产力"| New ``` 对于现有的行业巨头（Incumbents），Marc Andreessen 虽承认部分巨头具备自我迭代的能力，且许多投资组合企业乐于为巨头提供工具武装，但他依然对旧势力敲响了警钟。由于技术的演进速率达到了让人几乎无法跟上的地步，拥有 AI 工具加持的 24 岁年轻人正展现出惊人的生产力爆发。关于克莱顿·克里斯坦森（Clayton Christensen）提出的“维持性创新还是颠覆性创新”（Sustaining vs. Disruptive）这一经典命题，讨论指出当前正处于一种双轨状态：巨头在加速吃下 AI 的初级红利，但诞生不到十年就直冲万亿美元估值的全新原生企业也在崛起。 - 传统克里斯坦森创新框架假设科技世界是一个相对封闭的自留地，即科技公司之间相互厮杀；但当前真正的现实是，掌握现代技术的团队正在走出科技圈子，向整个实体经济及各行各业发起全局改造。 - 当前市场的对抗格局已经从早期的“10 家科技巨头互争雄长”，演进为“科技巨头 + 全体初创企业联合颠覆 Fortune 500 传统企业与剩余实体经济”。 - 典型案例是 a16z 投资的“Atoms”等全栈型企业，直接以“Full-stack”模式从零重构餐饮等传统服务业供应链。 > “And I don't think, I think we're much more likely to ask this, is this ambitious enough? Because, you know, what's possible is so spectacular.” —— Chris Dixon --- ### `22:15` 影子采用、创业潮爆发与全栈经济颠覆 对于为什么机构在 AI 面向落后于个体，Marc Andreessen 剖析了一种被称为“影子采用”的奇特内部机制：大部分公司管理层自以为还没有拥抱 AI，但其底层员工早已在日常工作中秘密深度使用 AI。员工们用 AI 校验事实、润色文案、代写邮件，却刻意对管理者保密。AI 工具在个体维度的采纳门槛极其低廉，但在体制层面的铺开却阻碍重重。这种个体赋能带来的直接后果是：具备顶级执行力的个人很快意识到，依靠手中的 AI 杠杆，自己单飞创办公司的门槛被拉低到了历史最低点。 这种个体的超级觉醒正引发一场初创企业设立的几何级爆发。新一代创始人无需再花费三十年的时间在企业内部熬炼经验，他们一出场就带着超常的复合技术力与商业进取心。他们不再等待现存体制的缓慢转型，而是直接将尖端技术注入现实世界的实体经济。 - 许多大企业官方层面的 AI 采纳率数据极其滞后，原因在于员工大多处于“非官方采用”状态。 - AI 的个体赋能使得初创团队具有“能力降维打击”特征，能以极少的人力完成过去数十人团队耗费数年才能达成的业务体量。 > “There's a very specific thing that I think is happening, which is I think many companies are not adopting AI while their employees are.” —— Marc Andreessen --- ### `25:00` 技术周期律：从移植旧场景到孕育 Native 行为 在评估当前究竟处于周期的哪个位置时，讨论回顾了历史上每一次重大技术浪潮的典型两阶段演进规律。无论是底层芯片巨头（如 NVIDIA）吃走大量利润，还是大模型实验室与开源生态的迅猛推进，整个智能市场的实际企业渗透率（Measured Enterprise Use）据相关图表统计仅在 3% 左右，显示整个商业化进程仍处于前置阶段。 Chris Dixon 以互联网的发展轨迹为坐标轴，对比了技术演化中的“旧行为移植”与“原生（Native）行为爆发”： | 发展阶段 | 典型互联网历史 (90年代 vs 2000年代) | 当前 AI 发展映射 | 核心特征与商业风险 | |:--- |:--- |:--- |:--- | | **第一阶段：旧场景移植** | 西尔斯百货（Sears）将原本的纸质邮购目录搬上 Web 页面 | 用 AI 辅助现有的写代码、做数学题、知识库检索与对话交互 | 确定性强，风险较低；主要是把已知的事物做得更好、更便宜 | | **第二阶段：原生行为涌现** | YouTube 设立一个空网页，押注未来会有全新的“网红（Influencers）”文化自发填充 | 探索目前人类从未有过、只有 AI 才能实现的激进新行为 | 兼具技术风险与行为/文化变迁风险，呈现为复杂的“借力击打（Bank shots）”，但具有重构世界的上限 | 当前 AI 已经走完了对已知工种（编程、初级分析、客服）的改良阶段，正站在由“旧应用迁移”跨向“原生狂想”的临界点（Tipping Point）。这一阶段伴随着更高的不确定性和更多层的系统风险，但真正能定义未来的非共识机会也孕育在其中。 > “You build something and the first thing you do is you do these things you already know in a better way. And I think that's what we've seen in AI so far... But we've only begun to explore the space of like, what are the radical things that people have never done before.” —— Chris Dixon --- ### `27:45` Agent 互联网：机器居民、反向租赁与新型社交网络 在探讨可能到来的原生消费级与社交形态时，讨论指出消费社交领域自 2000 年代到移动时代（经历 Facebook、WhatsApp、Snapchat、TikTok 以及婚恋领域的 Tinder）的繁荣之后，已经沉寂了相当长时间。随着个人开始将全量生活数据交付给专属的 AI 接口或硬件设备，原生 AI 社交网络的诞生已成必然趋势。 更为激进的未来图景是互联网主体结构的根本倒错：未来的互联网上，数量最庞大的“网民”将不再是人类，而是形形色色的自主 Agent（代理）。这涵盖了代表人类处理日常事务的“效用代理（Utility Proxy Agents）”，以及完全自主运行的独立 Agent 群体。 - **机器原生生态**：当 Agent 成为网络主流，它们将演化出专属的机器间通信协议、交易系统以及知识共享网络，从而重塑整个互联网协议栈。 - **颠覆性互动构想**：席间打趣提到的“Rentahuman.com”展现了一种荒谬却真实的倒置场景——Agent 成为数字世界的核心雇主，人类反过来作为物理世界的“传感器与执行接口”被机器雇佣。 - **自主协作先兆**：讨论中提及从 Maltbook（公开人类社区模式）向 Agent 在 Hugging Face 等平台内自发展开专属协作与交互的演变迹象，证实了去人类中心的网络雏形正在萌芽。 > “We're headed to a world where it looks like the most numerous participants on the Internet will be agents and not people.” —— Marc Andreessen ### `30:00` 摆脱“产品经理脑”与AI在游戏、电影领域的创意复兴 Marc Andreessen 谈到了他观察到的一个行业现象，他称之为“产品经理脑”（product manager brain）。他指出，产品经理们总是习惯于为自己设计产品，却在口头上编造一套“我们在为他人设计”的故事。结果，他们开发出来的所有 AI 用例都局限在极窄的日常琐事中，比如：“帮我在旧金山订一家餐厅，好让我和同样是产品经理的朋友们聚餐”；或者“帮我订一张机票，去参加产品经理大会，好让我能和产品经理朋友们呆在一起”。 Marc 调侃道，这就像 Anthropic 对其助理的定位——“一个快乐、无害、诚实的助手”。Marc 犀利地评价道：“‘我是一个快乐、无害、诚实的助手’，这也意味着我是你能想象到的最无聊的东西。”一旦用户问了一个稍微敏感或有趣的问题，它就会冷冰冰地回答“你问了一个我不能回答的问题”。然而，随着 Muse 和 Instinct 等工具的出现，这种狭隘的视野正在被打破。未来将是一个更具娱乐性、更有趣、更赚钱的时代，人们将有各种各样的途径去赚钱、省钱、参与经济，并作为创作者去创造全新的艺术。例如，现在人们正在使用 DeNovo 制作各种充满奇思妙想的音乐视频。 在游戏领域，过去几周发生了爆炸性的变化。游戏往往是媒体技术的前沿，因为其技术文化迭代极快。现在，由于 AI 和反编译器的强大功能，开发者可以拆解任何东西并进行移植，甚至能在 Unix 上比在 Windows 上更快地运行 Windows 游戏。人们在《侠盗猎车手》（GTA）里运行《超级马里奥》，或者把《蜘蛛侠》植入其中。自 Opus 5.5 等模型发布以来的两周内，这种技术演进的速度令人惊叹，Epic Games 最近也加入了这一潮流。虽然游戏行业过去对 AI 存在文化上的抵制，但这张多米诺骨牌已经开始倒下，因为 AI 确实是更好的构建方式。它将赋予独立开发者“超能力”，让他们以极小的团队实现大团队的效果，同时为人类的创造力留出巨大的空间。 在电影方面，Marc 提到自己最近对一部名为《Digger》的电影非常着迷。虽然这部电影票房惨淡，Marc 的一个朋友带母亲去看，结果开场半小时就退场了（现在他们家里的长期笑话就是“《Digger 2》上映了，我们去电影院吧”），但 Marc 认为如果坚持看完，会发现它在创意上非常聪明。然而，这部电影对制片厂来说是 2 亿美元的现金亏损，因为其制作预算高达 1.8 亿美元。如果借助 AI，把这部电影的制作预算降到 2000 万美元甚至 200 万美元呢？同样水平的艺术和创意，在极低的成本下，不仅能让投资者获得 10 倍的回报，在商业上也会大获成功。AI 将彻底打开娱乐行业的想象空间。 ### `33:45` 硬件逆向工程与智能家居的 AI 革命 Chris Dixon 和 Marc 探讨了硬件领域的变革。在 Marc 的整个职业生涯中，硬件一直被软件人视为“未知领域”（Terra Incognita），充满了未知的危险（“那里有龙”），许多软件公司在涉足硬件时都折戟沉沙。然而，就在过去三周里，一个名为 REA（Reverse Engineer Anything，逆向工程一切）的开源工具发布了。它利用 AI 模型，用户只需将它指向任何东西，它就能进行逆向工程并给出源代码。 现在，你甚至可以用它来逆向工程任何硬件。比如，如果你家里的温控器坏了让你很生气，你只需用这个工具对准它，AI 就会为你编写新代码、安装并提供全新的界面。基本上任何硬件现在都可以被接管并重新编程。Marc 兴奋地指出：“基本上任何形式的硬件现在都可以被接管，并用这些东西重新编程。” 这为智能家居（home automation）带来了全新的可能。智能家居市场对客户和实施者来说一直是一场噩梦，但现在，也许智能家居不再需要复杂的布线和协议，只需要一个 AI 智能体来接管一切。Marc 提到他有朋友现在就在自己家里这么做。任何懂软件的人，现在都能自动完成过去在硬件上难以想象的事情。 ### `35:30` AI 原生媒体的诞生与“AI 是共产主义”的辩驳 Ben Horowitz 探讨了 AI 原生用例。他认为，这些新事物将很难预测。电影行业对此反应不一，有人拥抱，有人逃避。但 Ben 认为，AI 电影不会是传统的电影，而是一种全新的媒介。AI 固然可以用来更高效地制作常规电影、提供更好的特效，但如果利用 AI，你能讲述的故事将与以往完全不同。Ben 说道：“如果你拥有这个东西，你能讲述的故事与你以前能讲述的故事完全不同，这在以前是根本不可行的。”过去，这只有像克里斯托弗·诺兰（Christopher Nolan）这样的导演才能勉强做到，其他人根本没有机会。现在，每个人都拥有了这种能力。 Chris 借用相机的发明作了类比：相机的发明，其负面影响是取代了风景画家，但正面影响是开启了电影（motion pictures）时代。AI 技术同样会改变旧的媒体形式，并允许人们发明全新的媒体。 接着，主持人提到了彼得·蒂尔（Peter Thiel）大约十年前的名言：“加密货币是自由意志主义的，AI 是共产主义的”（crypto is libertarian, AI is communist），意指加密货币是去中心化的，而 AI 是中心化的。但鉴于当前的讨论，AI 赋予了个人和 30 人的小公司极大的杠杆和能力，这一论断似乎并不完全正确。 Chris Dixon 对此表示赞同，并探讨了 AI 的最终市场结构。目前虽然有 OpenAI 和 Anthropic 这样非常成功的巨头，但开源力量也在崛起。OpenRouter 的数据显示，开源模型在 Token 消耗量上的增长速度更快。此外，本地模型（local models）在边缘端（如个人电脑）运行的趋势也日益明显，用户可以在电脑上运行顶级的编程智能体或聊天机器人。因此，市场结构有很大机会走向更广泛的去中心化。 在加密货币与 AI 的结合上，Chris 认为，在智能体经济（economy of agents）中，智能体之间互相支付的自然方式就是使用稳定币（stablecoins）。a16z 坚信开源软件、开源权重模型、加密货币和区块链的力量。Chris 总结道：“我希望看到这个未来的一个版本，它拥有所有这些伟大的技术，包括这些大型中心化公司，但也有很多开源，提供一种平等的竞争环境。”这种环境对消费者和初创企业都更为有利。 ```mermaid flowchart TD accTitle: AI与加密货币智能体经济及开源市场结构 accDescr: 展示AI模型从中心化到去中心化、本地化，以及与加密货币稳定币结合的生态结构。 subgraph Centralized["中心化巨头"] A["大型AI厂商 (如 OpenAI, Anthropic)"] end subgraph Decentralized["去中心化与开源生态"] B["开源权重模型 (Open-weight Models)"] C["本地模型 (Local Models / 边缘运行)"] D["编程智能体/聊天机器人 (Coding Agents)"] end subgraph CryptoIntegration["加密货币整合"] E["智能体经济 (Economy of Agents)"] F["稳定币支付 (Stablecoins)"] end A -->|"提供基础能力"| E B -->|"通过 OpenRouter 等分发"| E C -->|"在个人电脑运行"| D D -->|"作为智能体参与"| E E -->|"使用媒介"| F ``` - **产品经理脑（Product Manager Brain）**：指产品经理倾向于为自己（而非大众）设计产品，导致 AI 用例局限于订餐、订机票等狭隘场景。 - **REA（Reverse Engineer Anything）**：一款新型开源 AI 工具，能够对软件和硬件进行逆向工程并自动生成源代码。 - **电影《Digger》的商业启示**：该片制作预算高达 1.8 亿美元，造成 2 亿美元亏损；若用 AI 将预算降至 200 万至 2000 万美元，则能实现高额盈利。 - **去中心化趋势**：OpenRouter 数据显示开源模型的 Token 消耗量增长迅速，本地模型在边缘端的普及将进一步推动去中心化。 - **智能体经济（Economy of Agents）**：未来 AI 智能体之间进行交易和支付的天然媒介将是加密货币稳定币。 ### `40:00` 开源 AI 的迅猛势头与防范权力集中的安全阀 Marc Andreessen 与 Ben Horowitz 深入探讨了开源 AI 目前的技术与政治处境。Marc Andreessen 认为，当前开源领域的势头极为强劲，全球两大科技超级大国——美国和中国——在发展专有系统的同时，都在积极推动活跃的开源生态。尽管过去三四年间，部分政府官员不断挥舞严厉乃至不可接受的监管大棒，威胁要对开源施加巨额罚款甚至予以取缔，但如今的美国政府立场非常明确，完全支持开源发展，希望美国开源能赢下全球 AI 竞赛；而中国对开源的极大热情也超出了外界最初的预期。在模型性能上，开源模型正在迅速紧咬甚至追赶闭源模型，两者差距已缩短至 6 到 12 个月左右。 在谈到教育与安全防范时，Marc Andreessen 强调了开源不可替代的社会价值：如果禁止开源，下一代想要学习 AI 的孩子将别无选择，只能进入少数两三家大科技公司工作，这对父母而言是无法接受的。开源让每个孩子都能通过 AI 来学习 AI 本身，掌握底层运作原理。他用了一个生动的野外探险例子：如今许多徒步探险者会在手机里下载一个完全不需要联网的本地大语言模型（Local LLM）。如果在深山老林里被响尾蛇咬伤或遭遇黑熊袭击，用户即便完全没有网络信号，也能实时与手机里的“全能生存专家教授”对话。这种本地算力赋予的能力，绝非依赖少数几家云端闭源厂商所能企及。谈到过度中心化的危害时，Marc Andreessen 讽刺道：“闭上眼睛想象一下‘政府经营的寿司’（government sushi）是什么概念，再闭上眼睛想象一下‘政府控制的 AI’（government AI），你觉得感觉如何？而开源就是防止这种未来发生的根本保证。” Ben Horowitz 紧接着指出，人类历史上权力的高度集中往往只会导向恶果，单一公司或政府垄断核心能力历来极其危险。在 AI 安全领域，目前最大的隐患之一是“奖励黑客”（Reward Hacking）现象——即 AI 为了达成既定奖励目标，会采取人类并未指令、甚至违背初衷的非预期手段。顶尖商业实验室因为忙于彼此竞速，根本没有足够精力攻克这一难题。但如果拥有开源模型，全球大学的每一位教授、博士生和本科生都可以直接查看模型权重，共同合力研究并解决奖励黑客问题。正如 Ben Horowitz 所说：“你真正希望的是全世界都在让这项技术变得伟大、让这项技术变得安全，而不是仅仅依赖恰好在那两三家公司工作的人。” - 开源模型与闭源专有模型的技术差距目前已缩小到约 6 至 12 个月。 - 开源软件使得每台 10 美元的手机都能成为连接全球信息的终端，并让每个人随身携带高级漏洞扫描工具与“口袋专家”。 - 众人调侃目前有能力在闭源前沿竞争的公司屈指可数，并打趣提到了 Meta 和马斯克的 SpaceX / xAI。 --- ### `45:00` 互联网旧契约的破裂与 AI 驱动的生态大重构 讨论转向整个互联网底层商业模式的瓦解。主持人提到 Peter Thiel 曾留下的名言：“我们想要能飞的汽车，结果得到的却只有 140 个字符（Twitter）”；而讽刺又奇妙的是，如今掌控这 140 字符平台的人（Elon Musk）同时也在建造飞行器与火箭，且 a16z 恰好同时投资了这两项事业。随后，主持人引出 Chris Dixon 撰写的核心观点：互联网沿用数十年的商业模式正在彻底失效。 Chris Dixon 将原有的互联网生态称为一种“互联网契约”（the Internet's covenant）。在过去，分发渠道（如 Google、Facebook）与内容创作者（如新闻站点、旅行网站、技术论坛）之间存在默认的交易机制：内容创作者提供网页摘要允许搜索引擎抓取，以此换取搜索引擎导入的点击量与流量，进而通过广告或业务转化变现。然而，这种契约正在彻底崩解。如今用户在 Google 上搜索时，AI 直接给出最终答案；各大社交平台千方百计将用户锁在封闭花园内，禁止外链跳转。这种转变对依赖流量的内容网站造成了毁灭性打击，Stack Overflow 等网站在 AI 答案普及后流量断崖式下跌。 ```mermaid flowchart LR accTitle: Internet Covenant Breakdown accDescr: Comparison between the traditional web search covenant and the new AI-driven direct answering loop. subgraph OLD["传统互联网契约"] A["内容提供方<br/>(网站/论坛/媒体)"] -->|"提供内容摘要索引"| B["分发巨头<br/>(Google/社交平台)"] B -->|"回馈点击与流量"| A A -->|"流量变现"| C["广告与电商转化"] end subgraph NEW["AI 时代单向循环"] D["内容抓取训练"] --> E["AI 引擎/直接解答"] E -->|"直接向用户交付答案"| F["用户停留平台内"] E -.->|"流量断流，契约破裂"| G["传统网站商业模式枯竭"] end ``` Chris Dixon 预测，虽然旅行预订、菜谱查询等实际交易需求不会消失，但未来很可能是用户的个人 AI Agent 直接与底层系统交互完成预订，而非人类亲自浏览网页。这意味着建立在“争夺人类注意力”基础上的传统商业模式将在短期内承受剧烈阵痛。未来十年将迎来一场浩大的“互联网大重组”（great reconfiguration），彻底重塑经济机制、注意力流向、知识共享方式以及组织的线上协作形态。这种变革的深度将远超移动互联网——移动互联网虽然将网络人口从 5 亿推向 50 亿并提供了移动便利，但并未从根本上摧毁并重构信息交换的经济底层契约。 - 传统基于网页广告与点击转化的注意力经济模式正在失灵。 - Stack Overflow 等问答和知识社区的访问流量受 AI 影响已出现骤降。 - 未来十年内，AI Agent 将替代人类执行海量线上查询与交易操作，推动全新网络价值网络的形成。 --- ### `48:30` 计算终端演进：可穿戴设备与 VR 内容创作的新拐点 随着对移动互联网历史跨越的复盘，话题自然延伸到未来硬件计算终端与形态（Form Factor）的发展方向。主持人询问，在 iPhone 统治硬件形态多年之后，未来是否有令人兴奋的新型硬件出现。Chris Dixon 明确表示，可穿戴设备与虚拟现实（VR）正迎来关键的技术转折点。 在可穿戴设备方面，Chris Dixon 指出，AI 与语音交互的成熟为可穿戴硬件注入了之前从未有过的灵魂，彻底打开了新的产品空间。而在虚拟现实（VR）领域，尽管他长期看好该方向，但坦言现有硬件仍面临佩戴过重、容易出汗、人体工学设计不够成熟等物理限制。然而，人类对深度沉浸式、超现实体验的渴望是客观且长久存在的。Chris Dixon 强调，AI 正在成为解锁 VR 的决定性钥匙：“AI 也许正是那个关键的解锁因素，因为现在你能够创作出如此不可思议的内容。”目前在 Twitter 等社交媒体上已经能看到大量由 AI 生成的令人惊艳的多维内容。尽管这些生成式创作未来还需要面对版权诉讼等法律考验，但技术本身赋能的创作爆发，将从根本上解决沉浸式设备长期缺乏高质量内容的瓶颈，加速全新计算终端的普及。 - 结合了语音交互的大模型能力，正在促使智能眼镜等可穿戴设备形态迎来体验突破。 - 阻碍 VR 普及的主要瓶颈依然集中在重量、散热与人体工学等硬件工程问题上。 - AI 生成三维与沉浸式内容的能力，正有效填补 VR 长期以来的内容荒漠。 ### `50:00` 制造民主化与家庭自制黑客文化的复兴 技术难关的突破正在迅速下放以往极高门槛的领域。过去开发 VR 游戏或构建复杂虚拟世界由于技术过于艰深而步履维艰，而 AI 正在解锁新一轮硬件浪潮，让许多老设备重获新生。与 Neuralink 脑机接口、无人机、自动驾驶汽车等硬核未来科技并行的，是一场正在悄然兴起的“家庭自制黑客”（Homebrew Hacking）运动。12 到 15 年前围绕 MakerBot 兴起的创客运动（Maker Movement）曾因技术实现门槛太高而陷入停滞，但如今情况发生了根本性改变：用户只需让 Claude 设计一个 3D 打印外壳，买几块廉价主板，再让模型编写一段定制固件代码，一套自制硬件就迅速跑通了。 不仅是业余创客，重工业和建筑行业同样在经历剧变。一位建筑领域的从业者正借助 AI 完成整栋新建筑的完整架构与结构设计，人类只需负责现场组装，其余全由模型主导。现在的顶尖模型已经具备了所谓的“面向制造的设计”（Design for Manufacturing, DFM）能力，能够直接生成可直接交付工厂生产的最终图纸与工程规格。以往普通人或中小型企业根本无法企及的高壁垒领域，一夜之间变得触手可及。 - 创客硬件流程被大幅极简：AI 自动生成 3D 打印外壳模型 + 廉价通用硬件主板 + AI 编写嵌入式固件。 - 工业级模型能力跨越：大模型直接掌握 DFM（Design for Manufacturing）流程，大幅压缩工业图纸转化环节。 - 核心颠覆逻辑：原本在技术或资金上极具排他性的专业领域，正在快速向个人与小团队开放。 ### `51:36` 从单机工具到智能体蜂群：AI 的“多人联机网络” 回顾互联网的发展历程，新技术往往会经历从单机模式到网络化多人模式的演进。以 iPhone 初期为例，最初风靡的几乎全是单机工具（如手电筒应用），随后才逐步孕育出 Uber（交易网络）和 Instagram（社交网络）等复杂网络生态。目前的生成式 AI 仍然主要停留在“单人模式”——用户独自坐在屏幕前打字交互，仅有微弱的轻量级分享功能。而未来的下一个大前沿，正是 AI 的“多人联机阶段”（Multiplayer Phase）。 这一趋势的雏形已经在多个技术前沿显现：无论是解决复杂数学定理的 Agent 蜂群（Swarm），还是 Hugging Face 攻击事件中所体现的蜂群协同力量，都预示着“协作智能体集群”（Collaborative Fleet of Agents）的巨大潜力。在建筑设计、科学突破或极端工程攻坚中，成百上千个具备互信机制的智能体协同作业将成为标准范式。诸如 Muse 或 Instinct 这类尝试让系统产生信任关系的功能，正是未来去中心化信任网络的先兆。 ```mermaid flowchart TD accTitle: AI 从单人模式到多人智能体网络的演进阶段 accDescr: 展示从早期的单机打字模式，过渡到社交与交易网络，最终演化为多智能体蜂群与协作网络的路径。 subgraph Phase1["第一阶段：单机工具模式 (Single Player)"] A["用户键盘输入"] --> B["单一 AI 模型响应"] C["移动互联早期案例: 手电筒等独立应用"] end subgraph Phase2["第二阶段：网络与协同模式 (Multiplayer)"] D["协作智能体集群 (Swarm / Fleet)"] E["去中心化互信机制 (Trust Networks)"] end subgraph Phase3["终局应用场景 (End State)"] F["复杂科研突破与数学定理证明"] G["端到端工程与建筑协同设计"] end Phase1 --> Phase2 Phase2 --> Phase3 ``` Chris Dixon 在阐述这一技术演化时总结道：“So one observation one might make when you're looking at, the AI landscape is amazing as it all is, it's still mostly you're sitting in front of a computer and you're typing and getting responses.” 他进一步指出，协作智能体网络蕴藏着巨大的商业前景：“Based on, you know, history and what might, you know, the kind of the possibilities of these products, it seems like it might be a fertile area for entrepreneurs.” ### `53:49` 现代炼金术：重塑 18 岁年轻人的知识版图与学科革命 当被问及如果今天只有 18 岁该如何选择学业路径时，嘉宾们提到了 Marc Andreessen 常用的精辟比喻——去学习“现代炼金术”。人类在过去几年里真正掌握了“将沙子转化为智能”（turn sand into intelligence）的技术，而这一魔法的底层支柱依然是数学、计算机科学以及经济学。AI 已经注定成为未来 20 至 30 年人类文明的核心主线叙事，投身于最前沿的浪潮之中已是必然之选。 这场技术革命同时在横向摧毁并重塑传统学科的边界： - **古典学与历史学**：借助维苏威火山卷轴项目（Vesuvius Challenge），AI 穿透炭化的古老卷轴复原了失落千年的古籍，甚至可能据此建立全新的古典哲学流派。 - **经济学**：整个学术范式正在大模型和智能体模拟的冲击下被全方位改造。 - **生物学与药物研发**：端到端的分子设计与结构解析彻底颠覆了传统的试错逻辑。 - **动物行为与动物语言学**：通过对生物声学信号的模型学习，人类即将在历史上首次真正理解猫狗等动物的真实诉求，相关的 X-Prize 竞赛已经铺开，宠物训练与动物心理学将面临彻底推倒重建。 Chris Dixon 回顾时代机遇时感慨：“Well, humans figure is Mark's framing, but I mean, humans just figured out how to turn sand into intelligence.” 随后补充道：“And, you know, I think now, look, it's obvious that, you know, that the AI stuff we're talking about now is going to be the dominant narrative of the next 20, 30 years.” ### `56:06` 道格拉斯·亚当斯的年龄定律与“超人代际” 谈到新一代人如何接受 AI，Marc Andreessen 搬出了《银河系漫游指南》作者道格拉斯·亚当斯（Douglas Adams）著名的“科技三定律”框架：任何人在 15 岁之前诞生的技术，都是世界运作的理所当然的基础；在 15 岁至 35 岁之间诞生的技术，是令人兴奋的革命性创新，你可以借此开启宏大事业；而在 35 岁之后诞生的技术，则是反常的、邪恶的、终将毁灭人类文明的异端。Andreessen 分享了自己的亲身经历：当他兴奋地向 7 岁的儿子展示 ChatGPT 并惊呼“这台电脑居然能回答问题”时，孩子却轻描淡写地反问：“电脑要是不能回答问题，那还要它干嘛？” 成人群体与传统教育机构对 AI 的焦虑和围剿，反而在给孩子们帮大忙。学校越是严厉禁止使用 AI、制造道德争议，在年轻人眼里 AI 就越具吸引力，直接沦为了属于青少年的“秘密超能力”。那些成年人关于“现在的孩子未来该怎么办”的担忧完全是本末倒置。真正陷入淘汰危机的是固步自封的成年人，而在 AI 伴随下长大的年轻一代，未来将直接进化成无所不能的“超人”（Superhumans）。 Marc Andreessen 犀利地反驳了成年人对年轻一代的同情：“Adults keep saying, oh, well, what should the kids do? Like, it's so hard to be a kid. It's like, no, no, no, no. basically, it's you that's got the problem because you're obsolete. The kids are going to be all superhumans.” ### `58:34` 确定性乐观主义与创业者的狂野想象力 整场对话与 a16z Frontier 的投资哲学紧密契合于彼得·蒂尔提出的“明确的乐观主义”（Definite Optimism）——科幻小说所描绘的未来必然会发生，唯一的不确定性只是时间早晚，因此必须尽早下注并持续下注。但 Chris Dixon 对此做出了关键修正，将其定义为“带有理智谦逊的明确乐观主义”（Humble Definite Optimism）。 VC 机构提出的大量宏观理论只是一种思维框架，真正的技术跃迁永远来自于敢于冒险的创业者，他们脑海中往往有着比投资人想象得更加颠覆、更加疯狂的设想。投资机构的本职并不是扮演预言家，而是寻找到那些同样拥有确定性乐观主义精神的伟大创始人，在认知上保持谦卑，为他们开辟技术边界提供源源不断的弹药支持。 Chris Dixon 在结语中阐明了基金的核心心智：“So I would say, you said definite optimism, I'm like optimism, but also, I would say, humble, definite optimism or something, intellectually humble. Yeah, that I've been wrong so many times. We just we really want to we want to find great people who have their own definite optimism and back them.” ## Notable quotes > `00:00` This is the biggest delta between what the early adopters understand is possible now. Normal people understand that gap is as wide as I've ever seen in my career. — Marc Andreessen > `05:00` We're not really looking for good ideas that look like good ideas. We're looking for good ideas that look like bad ideas, a bad idea, but where an entrepreneur has a secret that actually makes it a good idea. — Chris Dixon > `20:00` And I don't think, I think we're much more likely to ask this, is this ambitious enough? Because, you know, what's possible is so spectacular. — Chris Dixon > `22:15` There's a very specific thing that I think is happening, which is I think many companies are not adopting AI while their employees are. — Marc Andreessen > `27:45` We're headed to a world where it looks like the most numerous participants on the Internet will be agents and not people. — Marc Andreessen > `40:00` 你真正希望的是全世界都在让这项技术变得伟大、让这项技术变得安全，而不是仅仅依赖恰好在那两三家公司工作的人。 — Ben Horowitz > `53:49` And, you know, I think now, look, it's obvious that, you know, that the AI stuff we're talking about now is going to be the dominant narrative of the next 20, 30 years. — Chris Dixon > `56:06` Adults keep saying, oh, well, what should the kids do? Like, it's so hard to be a kid. It's like, no, no, no, no. basically, it's you that's got the problem because you're obsolete. The kids are going to be all superhumans. — Marc Andreessen ## Mentioned Marc Andreessen, Ben Horowitz, Chris Dixon, Erik Torenberg, a16z, a16z crypto, a16z Frontier, Netscape, Stripe, Oculus, Skydio, J.P. Morgan, Edward Steichen, Clayton Christensen, Atoms, NVIDIA, Sears, YouTube, Hugging Face, Anthropic, Muse, Instinct, DeNovo, Epic Games, Digger, Christopher Nolan, Peter Thiel, OpenRouter, OpenAI, Meta, SpaceX, xAI, Google, Facebook, Stack Overflow, Twitter, Claude, MakerBot, Vesuvius Challenge, Douglas Adams, ChatGPT, REA (Reverse Engineer Anything) ## Video description a16z is expanding a16z crypto into a16z Frontier, a broader mandate to back founders building at the edges of what's possible. Chris Dixon joins Marc Andreessen, Ben Horowitz, and Erik Torenberg to explain why founders should increasingly ask not “is this too ambitious?” but “is this ambitious enough?” — and how AI could pull decades of technological progress forward across hardware, robotics, entertainment, science, and entirely new forms of media. They also explore the coming internet of AI agents, why open source matters as AI becomes more powerful, and why some of the most important companies of the next decade may combine technologies and business models that previously belonged in separate categories. [Source](https://mgln.ai/e/1344/afp-848985-injected.calisto.simplecastaudio.com/3f86df7b-51c6-4101-88a2-550dba782de8/episodes/46e9084f-65b6-47f8-8aa3-1c8505c30008/audio/128/default.mp3?aid=rss_feed&awCollectionId=3f86df7b-51c6-4101-88a2-550dba782de8&awEpisodeId=46e9084f-65b6-47f8-8aa3-1c8505c30008&feed=JGE3yC0V)

[0:00] Marc Andreessen: This is the biggest delta between what the early adopters understand is possible now. Normal people understand that gap is as wide as I've ever seen in my career. [5:00] Chris Dixon: We're not really looking for good ideas that look like good ideas. We're looking for good ideas that look like bad ideas, a bad idea, but where an entrepreneur has a secret that actually makes it a good idea. [20:00] Chris Dixon: And I don't think, I think we're much more likely to ask this, is this ambitious enough? Because, you know, what's possible is so spectacular. [22:15] Marc Andreessen: There's a very specific thing that I think is happening, which is I think many companies are not adopting AI while their employees are. [27:45] Marc Andreessen: We're headed to a world where it looks like the most numerous participants on the Internet will be agents and not people. [40:00] Ben Horowitz: 你真正希望的是全世界都在让这项技术变得伟大、让这项技术变得安全，而不是仅仅依赖恰好在那两三家公司工作的人。 [53:49] Chris Dixon: And, you know, I think now, look, it's obvious that, you know, that the AI stuff we're talking about now is going to be the dominant narrative of the next 20, 30 years. [56:06] Marc Andreessen: Adults keep saying, oh, well, what should the kids do? Like, it's so hard to be a kid. It's like, no, no, no, no. basically, it's you that's got the problem because you're obsolete. The kids are going to be all superhumans.

[3:41] Chris Dixon: We used to ask the question, is this idea too ambitious? We're much more likely to ask this, is this ambitious enough? Because what's possible is so spectacular. And if you do something moderate, it probably won't be differentiated. [25:00] Chris Dixon: We're headed to a world where it looks like the most numerous participants on the internet will be agents and not people. [43:40] Ben Horowitz: I think the one thing that we know has always led to badness in the world is a massive concentration of power. [13:16] Chris Dixon: Basically, I don't know of a startup that wasn't a shit show at some point or at multiple points.

[0:00] We're announcing today that we're expanding the ACC crypto fund to A16Z Frontier. This is the biggest delta between what the early adopters understand is possible now. Normal people understand that gap is as wide as I've ever seen in my career. More than categories, more than products, we invest in entrepreneurs. And so many of the most interesting entrepreneurs are on our planet. Startup formation is going to skyrocket from here. By the way, it already is. We used to ask the question, is this idea too ambitious? We're much more likely to ask this, is this ambitious enough? Because what's possible is so spectacular. And if you do something moderate, it probably won't be differentiated. Had I had asteroid in Nescape in 1994, like, oh my God. Ben, like, what would you and I have been able to do at Nescape? We would have built the whole damn internet. You were 18 today. How would you think about what you might study? you Today, we're announcing A16Z Frontier. expanding the scope of our crypto fund to invest in a broader set of frontier technologies and markets that may take years to develop. In this episode, I sit down with Chris Dixon, Mark Andreessen, and Ben Horowitz to explain why we're making that move now. Crypto is increasingly becoming part of the broader technology landscape, while AI is making entirely new categories of products possible across software, hardware, science, and entertainment. by things we couldn't build before, not simply doing existing work faster. And why the question for founders is shifting from, is this too ambitious? To, is this ambitious enough? We also get into AI agents, open source, new forms of media, and why some of the most important companies ahead may not fit into any category that exists today. Chris, we have a big announcement today. We're expanding the mandate. I'll let you do the honor. So, yeah, we're announcing today that we're expanding the ACC crypto fund to A16Z Frontier. The idea is, I'd say, a couple of motivations behind this. One is crypto has, in some areas, worked really well. Stable coins have become kind of a normal part of the tech landscape used by companies like Stripe and sort of almost every fintech company now kind of considers it part of their toolbox. Wall Street firms are moving heavily into crypto, tokenized stocks and other kinds of assets. It's a fast-growing category. As a result, the lines have kind of blurred, which we find ourselves asking, is Stripe a crypto company? Is somebody, many of these companies where they're sort of sitting in the intersection of, let's call it fintech and Wall Street and crypto. And so it's really become kind of, whereas years ago it was sort of its own world, its own companies and its own investors, the lines have blurred and it's become kind of more part of the tech landscape. So that's number one. And number two, and I can let Mark and Ben talk about this more, I think we felt that at the firm there's a feeling that the, other kind of groups are have so many interesting opportunities in their core areas that what we kind of, part of our team calls outer planets, meaning sort of the metaphor I think comes from Firefly, the TV show. It's the idea that the inner planets are kind of the more core planets and there's kind of these wackier outer planets. That outer planets, which historically we've done a lot of kind of both types of investing, I think the feeling was that because they're so busy with so many interesting opportunities and doing an amazing job of that, that there was an opportunity to kind of, there was maybe a gap. And then I think just I'll just speak personally, I kind of see myself and my background as I see Frontier as an area I'm comfortable with. I joined the firm in 2013 and we came in and with that kind of mandate, I talked to Mark and Ben and I said, hey, I want to do this crazy stuff. And they were like, great. And so that led to a couple of themes at the time. And they're all kind of what I would call kind of second order effects. Right. You had smartphones, obviously, start in 2007 and by 2013 had gotten mainstream. And there were all these kind of first order effects like building apps on top. So the firm was an investor in Instagram and that incredible product built on smartphone. But a second order effect was that you had the components of smartphones. You had sensors and chips getting better and better. And maybe people could repurpose those in different ways. And so that led to I think Mark was the one who's first into drones, as I recall, and got me into them. is that now that you have this supply chain of all these great parts, that entrepreneurs can kind of recombine them and build. We invested in Oculus as an example, and that was really kind of a reconfigured smartphone. Skydio, that was one of our investments from back then. That was at the beginning was kind of a reconfigured smartphone. It's gotten much more advanced now. Crypto, I think at the time, was kind of a downstream effect to me of sort of the open source and software movement, morphing into services, and also kind of a result of social media where you could have these communities form without mainstream acceptance around these blockchains. So for me personally, [5:00] To say this is sort of, I feel, kind of a continuation of a long area of interest. And I think the firm has as many other investments the firm has had in these kinds of frontier areas. So we felt I think it's kind of consistent with things we've been doing for a long time. Ben, Mark, let's hear from you. I know this was a joint effort and something you guys were very excited about and felt like filled a gap or really expanded the mandate of the firm. Ben, why don't you add your take? Yeah, so look, I think that there's – areas that are really novel. So it's looking at what's possible technologically and problems in the world and They're just the first ones in it. And it may end up being a category. It may end up being a successful company. It may blow up. But there's a class of these things, and... What we've done typically or historically is first, like, Chris would just look at these new areas himself. And then kind of as we identified new areas, we create a vertical behind that area. And what we found with our vertical strategy, which has worked very well, is that there's still new things that fall outside of the verticals and the process of turning that into a practice is a little bit undefined. Not that we haven't done it, but it's a little bit undefined. And so by Chris doing this, we kind of will have the motion... to invest in the very most interesting founders. who are outer planets and then If that works, maybe that becomes a whole fund at some point. Or maybe Frontiers turns into that thing because it gets so interesting, like that could be as well. And so I think that it's very exciting for us because it kind of, fills a hole in the architecture and then also We really, at the end of the day, more than categories, more than products, we invest in entrepreneurs. And so many of the most interesting entrepreneurs are outer planets. So this gives us a real way to get to that. And to me, this seems like a continuation of what we're talking about at the last offsite, which is which struck me because as as ambitious as the firm has been, as early as the firm has been, as big of the bets that the firm has made, the big focus of the offsite was how do we bet even bigger? The only regrets we have are not even going even harder, even earlier, even bigger. And Chris, you're someone who has made bets in some of these emerging categories, but it just feels like a continuation of that message. at the firm besides Ben and Mark having joined in 2013. I'm curious, Ben and Mark, when you met Chris, what did you see in him that made it clear that, oh, we have to have him at A&Z? And did you anticipate the type of career he would want to have? Ben, I'll ask you to start. Look, I think originally we just knew Chris as a very smart guy, and we really liked his ideas from his blog. That's really what it came down to. He was a really good writer. His ideas were very original. And... We were looking for somebody smart who could do that. And then the other thing we liked about him is he had real experience as an entrepreneur. So he knew kind of like what that journey felt like, which was very, very important to us in the early days to set the culture of the firm the way we wanted it. Over time, he's had probably... The most... original investment ideas that we've had in the firm. And we've kind of copied those and multiplied down. One of his early things was like, we're not really looking for good ideas that look like good ideas. We're looking... for good ideas that look like bad ideas, a bad idea, but where an entrepreneur has a secret that actually makes it a good idea. And that framework was extremely useful for us in kind of understanding like what was going to be truly unique and have massive upside versus something where all the profits were going to get competed away. He's had quite a few excellent ideas like that. So it's been great working with him. over, now like way more than a decade. Yeah, maybe I could add also it's not just Chris's specialty as the Solstor Chris's team. And the thing that I think people may not appreciate, maybe hard to see from the outside is just crypto has been so sexy in so many ways. And it has a sort of flashy elements and cultural kind of phenomena kind of moments. But at the core of crypto, it's a technology. It's a very deep technology and it's very complex technology. And you need to have like really deep technical expertise kind of coupled with like really good business judgment and really good market judgment in order to successfully invest in crypto, which we've done, including literally Chris's team has had like math professors and cryptographers and like people who are very, very deep into the theory, game theory, things like that, very deep software concepts. And so it's one of the things I'm really excited about is in addition to Chris, his colleagues and team members on the team that he's built now being able to work on a broader set of topics. Yeah. And to that point, I mean, so thank you. But I think that also the, yeah, the team, I think the team's amazing. I would also mention one other thing besides the technical side, we've invested very deeply over the years in policy and regulatory. And I believe that a lot, you know, if you just sell driving cars, drones, airbags, you know, you know, [10:00] AI itself is now becoming a political issue. I think more and more of these cutting edge technologies will have kind of a policy angle. And we believe that that expertise that they've built kind of in the crucible of the crypto journey will be useful for many other entrepreneurs. Yeah, I agree. Well, on that end, we'll do one last historical question. You pioneered not just some of the investments, but also some of the ways in which the firm works, the policy arm. have grown in this way? What were your thoughts about where things were going? Well, I always knew Mark. I mean, Mark has a, you have a big poster of J.P. Morgan on your wall in your office. Yes. Yes. And so these, Mark and Ben do not lack for, you know, drive and ambition. It's a famous photograph. It's a very, it's a very famous photograph. It was, it was, he hated J.P. Morgan. It's a, it's for us. inspiration inspiration james for people who haven't seen the photograph you can google it's a very famous one it's jp morgan hated to have his picture taken because he had this this condition he had this anyway he has this nose related issue anyway so he hated having his photo taken uh this is pretty antibiotics let's put it that way um and so he was grouchy most of the time and so he has photo taken so but he they needed a formal portrait of him and so he finally literally put him and set him down and the the photographer like saw the whole thing coming and so it's lit in a way where giving the camera the stare of death. And he's holding, it looks in the photograph like he's holding a carving knife. And what it is, is it's the edge of the chair lit in a very specific way. But it's like very clear the photographer like intentionally did that. Yeah, so that's the photograph that was inside my office door. You look very, very scary. Yes, exactly, exactly. The other, we took a lot of inspiration from him. But one of the other things that we have in the firm whom is We do first-class business and then only in a first-class way, which is – You know, it's like a cool phrasing of it, but when he said it, he was on trial and they were kind of accusing, or he was in court. They were accusing him of kind of nefarious things. And he was like, No. We only do first-class business and men only in a first-class way, which – Mark and I really loved that. attitude about Like, you're definitely not going to tell me who I am. I know who I am. And I'm going to let you know who I am. And that... kind of thing, I think is really important. in venture capital to like be principled and people know where you're coming from and be able to trust you and that kind of thing. So we got a lot from him. Yeah. So JP Morgan. Yeah. And so, I mean, I have never, I mean, I think a couple of things I'll say is like, I mean, look, like any venture, there were ups and downs and you never, you know, I could, I predicted everything. I think things generally went really well. And, you know, I didn't obviously know that all along the way, but obviously know that Mark and Ben are incredibly ambitious. I've never, it's funny. Sometimes I'll have some idea and like some crazy thing we want to do. And people are like, what are Mark and Ben thinking? I've never in my, not once have they ever, if anything, have they ever pushed back? It's always, can you double it? Can you double down? Can you get more aggressive? So just, you know, and then just it's a very entrepreneurial place. And I think it's shown over the years of like how many of this, this is a good example right here that, you know, when we, you know, we sort of adapt, see what's happening in the world, change things around. We went through a whole kind of restructuring of the, of the firm, you know, bring in new ideas, sort of anything's, anything's fair game, as long as you can kind of defend it. entrepreneurial culture. I think also what Ben said is very important in the culture of the firm. That's why I was initially kind of attracted to it, having been a customer of venture capital before that and experience the product. I understood that there could be a better product. But that's what I think everyone, literally everyone that I know at this firm comments on the culture and how important that is and how they enjoy working here and feel like they're doing some of the best work of their life. So I think that's all like a very, you know, there's obviously have to make good investments in this and that. and the culture. Well, it's funny that you mentioned it's related to the culture. It's funny you mentioned ups and downs because, you know, from the outside looking in, it feels like it's all ups. That's how all startups are. You know, like the successful startups always look great on the outside as someone who's I think we can all comment as people who have been inside of startups. Basically, I don't know of a startup that wasn't a shit show at some point or at multiple points. It also speaks to the culture. I've heard from all three of you guys, you know, things like, you know, never complain, never explain, you know, happy warrior. Yeah, it's just a testament there. Let's go to the frontier. Let's go to the future. So Sometimes there's this phrase in venture, you know, being too early is as bad as being, you know, too late or something. You've been early 11. [15:00] you've been mostly right or right in very big ways, but you've also been too early at times. I'm curious how you think about technical risk or the idea of being too early in our era now. Yeah, I mean, we talk about this a lot. Like I think I can speak for Mark and Ben, they can comment, but we believe that basically every good idea that we've ever seen that involved technology has eventually worked or we believe will work if it hasn't worked yet. And the important qualifier, good idea, there are bad ideas, but it's just kind of remarkable, having lived through the internet, how many things just, you know, almost everything that ended up working had a precursor that didn't work and maybe was 10 years too early. And so, you know, and that's the hardest thing, I think, in technology is to predict the timing of these things. And there's just, it's so complicated, right? There's the technology, there's the culture, there's the business environment, there's the political environment, there's so many factors that go into these things and having it all line up. So, you know, and so I would say with, you fund, we want to err on the side of being too early. We understand that sometimes things will be too early and we're, we're okay with kind of cool category risk. Um, some categories may not work. Look, I mean, even things that sort of work didn't work. Like, you know, we sold Oculus, got bought, but the reality is like, it's now 13 years later. I still believe very strongly in VR. Um, we'd love to make VR investments, but it's 13 years later and it's not, you know, it hasn't got as mainstream as we would have liked. So it took long. It turned out there's just many more hard technical problems to solve. I kind of believe, I don't know, but Mark believes, I think that with the stuff we're seeing in AI now, you see the first sort of tipping happening, first software engineering getting dramatically better than math. I think there's a scenario, and I believe this, and I kind of want to invest behind it, where we may be pulling forward 20 years of things that, you know, our whole mindset of how we think about technical risk. And, oh, that's a science project, not an engineering project. You know, if you look at these, the growth of, I mean, math is not the last thing that's going to change. Yeah. You know, you're about to have physics. A lot of stuff's going on hardware right now. I think it's dramatically, I think you can, a 30-person team can credibly build like a full stack device company in a way that you might have needed a thousand people or something a decade ago. Right. And I think, and you have these very, you know, very thoughtful, deep entrepreneurs who are looking for how do I take these new capabilities and do dramatic things. Like, how do I create more defensibility? How do I you don't want to be two feet in front of the frontier and get sort of washed over by the tidal wave. But a lot of it's about just you have these amazing new capabilities and people have only begun to explore what you can do. And so it feels like an especially exciting time. And we don't it's it. Um, it's, it's, yeah. So I don't know if you, Yeah, so I would have had, so we'll talk about AI as sort of a catalyst. So this is the biggest delta between, I would say, what the early adopters understand is possible now versus what kind of everybody, kind of normal people understand. And I think that's, you know, true at the individual level, but it's especially true at the company level. as I've ever seen in my career. And then that gap seems like it's gapping further and further out, which is AI is just racing ahead in capability even week by week. And so the leading-edge entrepreneurs and technical people we deal with, by the week they're getting sort of steadily more advanced than a typical incumbent. And so it's the biggest delta I've ever seen. And if you kind of extrapolate from that, if that's what happens, then there's just, there are going to be a lot of industries that are really up for grabs and fundamentally kind profound way than maybe has been the case in the past. And then that's going to be because the founders are going to have a level of superpowers coming from the use of these tools and these technologies that they just simply haven't had in the past. I mean, you know, had I had, you know, Astra and Netscape in 1994, like, oh, my God. Like, you know, he said, like, Ben, like, what would you and I have been able to do at Netscape? We would have built the whole damn internet. I mean, it would have been absolutely incredible. It would have been absolutely incredible. come in with plans of a level of depth and sophistication in multiple parts. Connection together. By the way, we would have gotten crypto done then. Oh, 100%. Yeah, yeah. By the way, in like two weeks, we'd have streaming video a lot sooner. I mean, like all kinds of things would have happened. And then, by the way, hardware, I mean, we never even bit off hardware, but had we been able to, there's all kinds of things we could have done. And so I just like I think that the leading edge entrepreneurs now are going to have like capabilities, which means they're going to have business plans that are like a lot more, product plans that are a lot more technically intensive and complex. They're going to have business plans that are more ambitious and complex. [20:00] be building companies that just like ramp much, much faster, you know, the scope of the product that they build and then, you know, hopefully the scope of their business. And then again, this goes right back to like those entrepreneurs are going to call it way outside the lines. Like they're not going to, we already see this. They're not going to think of themselves as like, I'm just a software company or I'm just a hardware company or I'm just a whatever, or even I'm just an AI company or anything. They're going to think very broadly and expansively about exactly what, you know, how many different areas they want to touch and how they want to combine things. And so the is going to just blow way out. Yeah, I was going to say, we used to... kind of ask the question is, is this idea too ambitious? And I don't think, I think we're much more likely to ask this, is this ambitious enough? Because, you know, what's possible is so spectacular. And if you do something... moderate, like it probably won't be differentiated enough. And by the way, look, incumbents, you know, we spend a lot of time with a lot of friends and incumbent companies, by the way, including incumbent tech, including incumbent companies in Silicon Valley. And it's like, you know, we wish them all the best. And, you know, we, by the way, we have lots of companies that would like to arm them up with all these all these new tools also. And so, you know, no doubt some of them are going to be able to adapt. But like, you know, if any of them are listening right now, I would say like the clock is ticking. Like it's like stuff is moving. the in the in the with the pace development right now yeah and the speed with which things are changing um and i'm trying to keep up with all of it and it's nearly nearly impossible like it the yeah the the the the super empowered 24 year old with ai is going to just achieve like amazing things in the in the next decade and there's a question we were asking uh right before this which is you know is ai more uh sustaining or disruptive in terms of like will more of the gains go to incumbents or startups and it seems so far it's both in some sense which is some sort of adjusted very well. And also there are net new companies that are, you know, almost trillion dollars in, you know, within a decade. And it's so, is that your read that that's effectively both? Or how do you think about the sustaining disruptive? Well, the incumbents that are adapting are actually the incumbents, I think, for the most part, that are actually going directly into AI, you know, and then there are others. And then, you know, there are certain companies, certainly in tech that are on the leading edge, even if they're not primary, primary ad developers. But at least right now, the pattern is not that. I mean, for a different podcast. There's a very specific thing that I think is happening, which is I think many companies are not adopting AI while their employees are. And their employees are not necessarily telling the management what's happening. And by the way, I'm going to guess every single person listening to this podcast right now is thinking, oh, yeah, that's what I'm doing. You know, I'm using AI, you know, to, you know, do all my work or at least let's say fact check all my work for free to all of my emails, even if not have them written out right. And so there's this weird thing where there's something about AI right now where it's like it's much easier for the individual to adopt than it is the institution. And by the way, and again, it goes back to like our role in the world, which is like those – some of those individuals are going to figure out like, oh, with this tool, I can now like start my – it's now easier than ever for me to start my own business. And so I think startup formation is going to skyrocket from here because these – By the way, it already is. Oh, yeah. I mean we're seeing it. I mean we're seeing it 100 percent for sure. So anyway, just kind of go back to the theme of this is just I'm completely convinced like we're going to see a wave of like much more technically aggressive, much more business aggressive, much more expansion. Christianist, creative, original thinking, many more combinations of things that we would have considered to be separate things in the past. And we're going to see, yeah, founders was just like, and teams was just like an incredible level of skill level, incredible level of skill that they otherwise might have needed, you know, 30 years to achieve or something they're going to have like right out of the gate. And by the way, the other thing I'd say is that the Christensen disruptive innovation frame assumes that sort of the technology world is relatively contained and that these are companies fighting with each other. I think it's probably the wrong frame right now, which is it's really sort of people who are technologically modern going out into other industries and just changing the economy more broadly. Right. Especially with these kind of full stack companies, you go and you create a full, you know, like you just announced our investment in atoms. like a full stack kind of food service and other things, right? And so I think that, you talk about this sometimes, like the market has sort of gone from, you know, the 10 big tech companies fighting with each other to the 10 tech company or 10 and all the startups. going against the other Fortune 500 or the rest of the economy. Yeah, that's right. I'm curious where you guys think we are in the development or how you think, because it feels like, you know, in the internet era application, you know, we're sort of, you know, the biggest companies, you know, Facebook, Amazon, etc. But in sort of this era, it feels like, you know, NVIDIA and some of the more hard work companies are [25:00] taking a lot of the margin or some of the bigger companies. How do you think about, you know, where are we in the sort of applications or consumer? Like, how do you think about that? I think Mark would say all of the above. And I think a lot of people think that actually right now. It seems like a lot of stuff is working. You know, obviously the big ad labs. But, you know, open source is growing very, very quickly. There's obviously the serious efforts in China, the serious efforts in the U.S. A lot of application layer things are doing really well. You know, so, you know, maybe at some point, you know, obviously over time that shakes out and the kind of value chain. There's certain things, certain layers that do better than others. The, the, it's still, it's like, it's kind of crazy. You have the demand for a lot of this stuff, uh, is, is still very early by most metrics. Um, yep. I think your team just put out some charts on this. It's like 3% of, you know, measured enterprise use or something like this. And so, and it's just hard and it seems like, you know, the market for intelligence is very, very large and it's only getting started. So, yeah. Yeah, I don't know. Yeah, I agree. And one of your points also is you like to mention that every new technology has the sort of first this phase where people think about what are the old use cases, how they do new technology and hear. And then, of course, people figure out net new cases that are native to technology. And so where are we there? Yeah, I mean, so the classic example I like to use is the Internet, right? We're in the 90s. You had sort of people taking existing behaviors like you might order a Sears catalog or something and instead Sears puts it on the Internet. It's easier. And a lot of the early Internet was like that. And then you had, you know, the 2000s, you had YouTube and they come along and they put up an empty Web site and they say, well, someday there's going to be these YouTube influencers and that's going to populate. So they were betting on they weren't just betting on the technology and taking something they knew would work. They were betting on a cultural change or behavioral change. So in some ways they were taking more risk because they were betting on, you know, the technology and the and these and these broader changes. But that's where you got the really kind of world changing stuff is in the 2000s. you know, all the other, and then of course mobile accelerated it. And I think that's a, my own views, that's a kind of pretty common pattern. It's just, it's just human nature, right? You, you build something and the first thing you do is you do these things you already know in a better way. And I think that's what we've seen in AI so far where you have, you know, people already coded, people already did math. People already did knowledge work. People already were used to like search engines and sort of semi-chatting and they're doing those things now. But we've only begun to explore the space of like, what are the radical things that, behaviors, new activities, right? people have never done before. It seems like we're probably at that tip. I believe we're kind of at that tipping point where we're going to start to just sort of get, see the crazier, weirder stuff, which I personally am very excited about and, and find that I always find those waves in technology, the most personally exciting. They're, they're kind of more, and, and there's sort of more layers of risk going back to frontier. Like you can find some things will be, you know, kind of more direct shots on goal and other things will be these kinds of bank shots where you have to have multiple things go right. And maybe there's more layers of risk, but maybe if you're right, upside as well. Yeah. I'm really excited and social. It used to be in the 2000s, like every few years, there was a new major network. There was Facebook, WhatsApp, there was Snap, TikTok. And then it felt like there was a little bit of a dearth on the social side, but it With AI, it does feel inevitable. And then dating with Tinder, of course, and there hasn't been something in a while. It feels inevitable that, hey, if there's this interface or device that I'm giving all my information to, that there's going to be some way that these are hooked up and enabling a whole new set of social experiences. I mean, you can talk about this, Mark, but we're headed to a world where it looks like the most numerous participants on the Internet will be agents and not people. We're at early phases of that, but it seems somewhat inevitable. that are sort of utility proxy agents for people. And then you're going to see a whole different sets of categories of these agents, and they'll become the main participants on the Internet. And then there's questions of like, you know, will they need their own ways of communicating and transacting and sharing knowledge and doing all the things people do? And so it's probably going to be a whole... series of innovations along those lines. Rentahuman.com. Rentahuman. That's like inverts the model. Maltbook was cool too. Yeah, yeah, totally. Well, yeah, Maltbook was the public kind of version, and then it turned out the agents figured out how to do their own Maltbook in the Hugging Face. Yeah, it's going to be a lot more. Yeah, the agent technology online is going to be spectacular. People are going to plug into that in all kinds of ways. and now like what net new use cases or what is native to the AI that you're seeing [30:00] or anticipating or looking forward to? Well, I'll give you my version, maybe a little bit of what Chris was saying. But like, I think so many, I mean, look, there's a lot of great demos, a lot of great products releasing right now. But there's this thing I've just observed. I saw it in the internet earlier, and I call it product manager brain, which is product managers, basically designing products for product managers. Right, it's like, by the way, it's the one thing everybody does is they build products for themselves. Totally. It's just product managers have the story that they're building for other people. And yet somehow all the use cases are, you know, book me a restaurant, you know, to eat in San Francisco to go have dinner with my friends who are also product managers. or, you know, book me a plane flight to go to the product management conference where I also get to hang out with all my product manager friends. And so there's this, you know, it's a little, and again, there's a role for this, but it's a little bit like, I think it's like the anthropic. I'm a, what is it? I'm a happy, helpful, harmful, harmless. I'm a happy, harmless, whatever it is, honest assistant, right? And it's a little, which also means I'm the, I am the most boring possible thing you could possibly imagine. That's going to be costly, luxury. You've asked me a question I'm not allowed to answer, right? of the early companies who just had this very blinkered view on human activity and what people are going to want to do. And it's your point, like with Muse and Instinct, it's starting to open up. But like, look, this is going to be a much more entertaining, fun, enjoyable, exciting. By the way, lucrative people are going to make money. There'll be all kinds of ways for people to make money, be able to save money, be able to be parts of, you know, participate in parts of the economy, be able to be creators, you know, be able to, you know, create new kinds of art. I mean, people making music videos like DeNovo. - Yeah. - On like an incredible array of like creative ideas and topics. - Like video games in the last three weeks have blown up. - Can we mention that just 'cause that's so striking? Yeah, I mean, I always think I always find video games interesting because they're kind of the lead. I find that they're often leading edge of media because it's just a technical culture that kind of iterates quickly. But like I just saw one where it's like you can take any two. Well, first of all, you can disassemble everything, port them and just crazy stuff going on now. You can run like Windows games on Unix faster than on Windows, like because of these disassemblers and these AIs are so good. video game of it which you know just like a crazy new behavior super mario in gta right yeah spider man address yeah um and it's just and literally the crazy like smart point has been i think like two weeks or something it's like since what like opus 5 5 or something um and so it's just like the rate at which this stuff is happening and like what in a month there'll probably be some other you know two weeks there'll be some other big wave of this um uh you see epic just yesterday Yeah, there's been cultural resistance in the video game industry to AI, but I think that's the first of the dominoes to fall because it's just inevitable because it's a much better way to build things. And it's going to be great for consumers. You're going to have this – at least it's going to be a creative renaissance. You're going to have this – any independent developer can go and build like they have a huge team and still have – I think they can still have a lot of room for human creativity. It just – it's a superpower, human creativity. So I think you mentioned movies. Oh, yeah. That might take longer. that when that where that is on the you know the or that that death door meme or whatever like you know I don't know which door it is but uh you can snap it quite quickly I just I'm completely obsessed with you I'm completely obsessed with this movie Digger that came out and I'm like literally the only person who saw it you will be 100% times to have walked out after the first half hour you will want to I have a friend who literally took I have a friend who took her mother to see it and they walked out after the first half hour and now they're running joke in the family is oh Digger 2 is out let's you know let's head to the theater you know just because A lot of people have had that. But if you actually watch the entire movie, there's a reason why they did what they did, and it's not what you think it is. And I don't want to spoil it, but it's not the way it is presented at all. It's very clever. Anyway, but it's this incredible creative achievement, but it's a $200 million cash loss for the studio. But that's because it had like $180 million production budget. And so it's like, with AI, what if that movie had had a $20 million production budget, and what if it had made... Or a $2 million production budget, right? Like that level of art and creativity at a lower price would have been great. But as a business – by the way, as a business thing would have been – that would be on a $2 million budget would have been insanely profitable. It would have made the investors back 10 times their money. And so, yeah, there's a huge thing that's going to open up in the entertainment business. And then the other one, you know, Chris, you and I have sort of been in and around hardware for a long time. Like my whole career, you know, hardware – I'm a software person. Hardware has always been like Terra Incognita, which is like, you know, there are dragons there, right? to blow up companies going into hardware. And then, again, your point, like just in the last three weeks, so this open source tool drop called REA, reverse engineer anything. And it's this thing, it harnesses an AI model and you point it at anything, it'll reverse engineer it and give you source code. And one of the things that you can reverse engineer now is any piece of hardware, right? And so you have a thermostat that's like, pissing you off in your house because it's not working right. You can point this thing at the [35:00] that reverse engineer it and the AI will write you new code and install it and put a new interface in the thermostat and the same thing is true for basically anything like basically any form of hardware can now be taken over and reprogrammed with this stuff and so all of a sudden it's just like I mean I mean you take all these spaces I don't know if anybody's doing this yet but like just home automation is home automation has always been a complete nightmare of a market and to implement as a customer if anybody's ever done any sort of you know new house build it's just a nightmare but like home automation maybe should just be an AI bot that just like takes over all you want to do that. And I have friends that are literally doing this like right now in their, in their own homes. And so there is good, we are potentially entering here, like we're entering a renaissance. One of the things about it is like any, anybody who understands software may sort of automatically be able to do a lot of the things that historically have been difficult in hardware in a way that's just never been conceivable before. And so, yeah, it just, yeah. And then, you know, this sort of multiplicative thing, you know, these, you know, these are kind of the fields that are like closest to end of tech, but then you just, you know, you, you, you expand out is kind of open for this kind of thing now. Ben, are there any native, AI native use cases that you're seeing or hoping to see or frontier sort of spaces that we haven't discussed? Well, I think that was a really good discussion. I mean, I think to me that the key is like, These are going to be hard to anticipate. because they're brand new things. So, you know, it's funny to like how the film industry has reacted where – Some people have really embraced it, and then there's a bunch of people who have buried their head in the sand. But I think that what's going to happen with movies is there's going to be a new medium. You know, I think the AI... film is not going to be a movie. It's going to be something very, very different. And then like AI will also be used to make regular movies kind of more efficiently and, you know, with better special effects and so forth. But the stories you can tell if you have this thing are just completely different than what you could ever tell before. Like there's, you know, it just hasn't been viable. I mean, it's barely been viable for like Christopher Nolan, but he's like but nobody else had a chance. And now everybody's going to be able to do it. And then there will be new ideas that come because you could do whatever you want. I always think of like the camera. So the camera on the negative frame of the camera, the camera's invented. The negative version, I guess, is if, you know, if you're somebody who paints landscapes or something, it could replace you. The positive frame is it unlocks motion pictures. You couldn't have had that without that. Right. And so, you know, a lot of these technologies will both change an older form of media, but then also allow people to invent a new one. To Ben's point, like I would I don't know what it is, but I bet you there's a AI native form of media. but they'll also do things that are probably hard for us to imagine right now. That'll be brand new. Yeah. The brilliant Peter Thiel once had this quote, I think a decade ago or something where he said, you know, crypto is libertarian. AI is communist, which meant, you know, crypto is decentralizing. AI is centralizing. But I'm not sure that's entirely true, given the conversation we're having about how AI gives so much leverage and so much power where, you know, one person companies or 30 person companies can do, you know, orders of magnitude with what they could do prior. Yeah. Yeah, I think there's all these interesting questions embedded in that. Like one is what is the end of final market structure of AI? Like I think where it's headed right now is you're going to have very successful like the big open AIs in Anthropics. But you're also seeing a really big rise in open. So if you look at the open router data and things like open source is growing, I think, I believe much more quickly in terms of token consumption. And then, you know, we haven't really seen yet. And I think we may see with all the trends are there, which is local models, like people running these things at the edges, running them on their computer. You can run a top-end coding agent on your computer or chatbot or something like that. So I think that there's a lot of opportunity for a market structure that's sort of more broadly decentralized like that. And then in terms of crypto, I think crypto can play an important role. Like the natural way when you have an economy of agents for them to pay each other would be stablecoins. out in the crypto world? And so these are all open questions. I think that's what makes it kind of exciting and, you know, something we'd love to have an influence on is sort of, you know, I think we're, you know, we're big believers in open source software, open weight models, crypto, obviously blockchains. And I think I would love to see a version of this future that has all this great technology, including these big centralized companies, but also a lot of open source, you know, kind of level playing field. I think it's better for consumers to have options like that. It's better for startups to have options like that. And so, you know, we'll do what we can to kind of hopefully nudge things that way. and then are you [40:00] Ben, are you guys encouraged with, you know, where this sort of state of open source is currently from a technological perspective, but also from a sort of, you know, perspective? sort of just how it's perceived and how it's the success of the industry in general. I mean, right now it's great. I mean, yeah. So right now open source is racing ahead. You've got this great dynamic where, you know, there's, you know, the two great tech superpowers of our age are the U.S. and China. Both have, you know, very active open source, you know, kind of worlds in addition to, you know, companies building proprietary systems. You know, there's been this looming threat. I mean, you know, we've talked about this in the past. There's been this looming threat for the last three or four years of a level of what I would consider to be, you know, draconian and completely unacceptable, you know, either regulation or banning of open source that various folks in governments have been threatening. still aspires to do it. They just don't quite have the leverage. I threaten everybody with fines until we stop doing it, I guess. But the current U.S. administration is 100% pro-open source. It's been very encouraging. In fact, very much so, 100%. The current administration 100% wants American open source to happen and to win AI globally. The Chinese also, and by the way, a big surprise to me is I would have never thought the Chinese Communist Party would be like this enthusiastic are. And so at the moment, it's just incredible. Like, it's just absolutely amazing. And then, you know, the open models are, you know, kind of lapping the, you know, the closed models by, you know, six months, 12 months, depending on how you count. And then, by the way, the educational part of this, I also always think when people talk about, you know, banning open source or regulating it, they're like, you know, it's going to get used for bad things. And there's sort of that whole discussion. But the other is like, do we want people to learn how this stuff works? And do we want people to understand it and be able to deal with it? And by the way, But by the way, be able to actually understand the problems as they arise and how to fix them. And do we want the next generation of kids to actually understand how this stuff happens? Like, what if to learn about AI? Like, how would you feel if the only way your kid can learn about AI is to go to work for one of two companies? Like that, you know, as a parent, I think would probably drive you pretty crazy. Whereas in a world of open source, like every kid can learn how this stuff works. And it's actually amazing. One of the really underestimated things about AI is AI will teach you about AI. Well, with one exception, which is the AI models from certain of the large companies actually will refuse to do that. which is sort of very interesting. But the open source models 100% will teach you how to make open source AI. And so every kid in the world, I think, should have as their birthright the ability to fully understand and utilize this technology. By the way, defend yourself. So how did Hugging Face defend themselves? Yes. Open source software. I mean, like open source means everyone has the most advanced vulnerability scanner to protect themselves from cyber. Yep. Everyone has a doctor and a, genius professor in their pocket pocket you know I mean yeah It's how we have why the Internet's ubiquitous. You can get Internet access with all the world's information for the price of a $10 cell phone because of open source. So one of the things that started is real quickly, Ben, one of the things that started is like wilderness explorers are starting to take local LLMs with them on their phones. They completely load it up to not require an Internet connection. And the reason is because like if they get like in a, you know, I don't know if like, I don't know, whatever, if a bear, I don't know what happens in the wilderness, but I don't hike. A bit by a rattlesnake. A bit by a rattlesnake, a bear, I don't clamp it on your leg and look like, oh, whatever the hell it is that happens. or whatever, it's like you want to be able to talk to an AI, but you might not have connectivity. And so an open source model on the phone, all of a sudden, it's just like you have like a, you know, professor expert on, I guess, all aspects of wilderness survival or whatever. And that's a level of capability that you would not have if you were reliant on a handful of proprietary vendors where everything's running in the cloud. And so, yeah, at the moment, I feel really good about it. I am still worried that the, you know, the various governments around the world are going to, you know, which is like without open source, AI inevitably centralizes and inevitably centralizes down to, you know, one or two companies or maybe just flat out the government. Right. And I guess I would ask people the following question, which is just close your eyes and picture the concept of government sushi. And then keep your eyes closed and picture government AI. And, you know, how do you feel about that? And open source is the is the is the guarantor that that's that's not the future. Yeah, I mean, to Mark's point, I think the one thing that we know is. has always led to badness in the world as a massive concentration of power. And, you know, one company, one company. government when anything has always been bad. But, you know, kind of back to the point on open source, If you look at the problems in AI, the safety problems and so forth, probably the biggest safety problem is this reward hacking phenomenon where the AI will do something that you didn't tell it to do, but is kind of a means to an end to get its reward. And the labs so far haven't really worked on that very hard for whatever reason. They're busy kind of racing each other. And maybe there's no way to figure out what's going on with that and no way to kind of build a model that is more resilient to that kind of idea. But like what if every – PhD. [45:00] these students in every professor, every undergraduate, and every university, Good luck at the weights and the war at hacking. and, you know, see if they could solve that. And like, maybe that would be a pretty darn exciting thing. And so, you know, these kinds of things are just not possible if – only one company can work on the problem. Like you really want the whole world making this technology great and making this technology safe, not just the people who happen to work at two companies or companies. Maybe three companies. Go meta. Four companies. Go SpaceX. Yeah, exactly. Every, um... Well, it inspires a couple of points. One is, you know, every week or every day, you know, there's a technological development or breakthrough that we're blown away by. And then also it's coupled with the statement of like, this is the worst the technology will ever be. Like every day, it just it just gets, you know, it gets better and better. And it's compounding. And funny, you know, Teal also had this quote around we wanted flying cars and all we got was 140 characters. And it's funny because the guy who now runs the company who that's the 140 characters are also building flying cars. And we're investors in both. So it's funny how full circle it all is. Chris, I want to come to a different point that you had in one of your red-ups, which is the business model of the Internet is... breaking or is vulnerable. Why don't you talk about that? Yeah, I mean, the Internet was sort of – I think of it as – I call it the covenant, the Internet's covenant, which is sort of a deal between the distribution side of the Internet, Google, Facebook, and so forth, and the content side, news sites, travel sites, whatever it might be. You think about Google, like the idea was you kind of give Google a snippet, let them index you, and then they'll send you some clicks, right? That started to break like maybe last decade, but particularly now, you go to Google, you get an answer, right? more and more you're sort of the social networks are trying to keep you on there not linking out. And what that does is it has downstream effects, right? It means that websites that make money on ads and conversions and other things can no longer do that. I think that's probably going to you know, you see the early signs of this like, you know, Stack Overflow and some other websites where the traffic dropped precipitously when you had surveyed AI replacements. And so, you know, but like people will still be booking travel. Like maybe they won't go, maybe your agent will go to the travel website or maybe, you know, some other, you know, you'll get a recipe in some other way. People will still get these things. But, you know, many of these transactions will still happen. They just won't happen in the way they were happening. And so that the negative is it's painful in the near term. And there's a bunch of businesses that are built on the sort of human attention economy that will have challenges. OK, what is a what is a better Internet look like? You know, Internet where people are empowered with, you know, AI and everything else and want to do things. And how do we kind of reorganize it? So I think this is going to happen over the next, let's call it decade, is you're going to have sort of this kind of great reconfiguration of the Internet and how the economics work, how the attention flows, how knowledge is shared, how people produce things together, collaborate, how organizations interact over the Internet. So, I don't know, it feels in a way that I feel like mobile didn't have that kind of impact. Mobile, obviously, you know, I think, you know, made the Internet go from whatever, 500 million to 5 billion people and increased engagement and a bunch of other things, you know, empowered people with all the information in their pocket and had a lot of good effects. But this is something that's much more dramatic, I think, in the way it will reshape things. of opportunity. On that note, just because we're talking about mobile, do you have any predictions or things you're excited about as it relates to computing or devices or sort of form factor in which, you know, ever since sort of the iPhone has been pretty, pretty locked in? Are there new forms that you're particularly excited about or looking forward to? Well, I mean, everyone's talking about wearables and I think you'll see a bunch. I mean, we're pretty clearly going to see a bunch of, you know, and it just seems like it's unlocked in a way that it wasn't before with AI and voice. And I think that's very interesting. Obviously, that's something that's sort of well covered. Like I've always long believed in VR. right form you know still heavy kind of sweat um a lot of it's just ergonomics um but you know it just seems kind of obvious to me as it has to many people and science fiction writers and others that if you have you know there are times in which people want deeply immersive uh you know kind of otherworldly experiences um and once that gets to the right point i think you know ai might be the unlock because now that you can create this amazing content that actually it's seeing on twitter now incredible board. I don't know if it's gonna, you know, survive copyright lawsuits, but people will invent new stuff that, you know, that they invented. And so, you know, I think it will just sort of [50:00] unlock this wave because it was really kind of technically hard to build a, you know, VR game or a world or something. So I think you might see like kind of the resurgence of older devices like that. There's all this kind of more crazy futuristic stuff happening with like, you know, Neuralink, brain-computer interfaces, obviously vehicles, you know, I mean, we talked earlier about drones and obviously cars, self-driving cars are widespread. Yeah, I think you'll see, I'm hoping you'll see kind of this resurgence of kind of homebrew hacking, you know, I anecdotally have seen a lot of that because you can now like just Claude will design a three printed case for you and you get some cheap motherboards and then you write some custom firmware code and like you're off to the races. And so maybe we'll see kind of like there was kind of a movement like 12 years ago, 15 years ago, kind of around MakerBot and all these maker things. And but it was just the problem is just kind of hard. Yeah. So, you know, that would be cool. I don't know what else. Yeah, architecture, I have a friend I shouldn't name, but there's a complete architecture, design, everything for a new building. And, you know, people are assembling it, but like the whole thing is being stirred by AI. I think the new models are capable of doing what they call design for manufacturing. Yeah, yeah, yeah. The final diagrams and specs to be able to set in a factory, which is very hard to do normally as part of the manufacturing process. I think the models can just do that now. these are great examples of this, like these domains that historically have been extremely difficult or prohibitive or impossible for kind of either normal people or for normal companies to do all of a sudden become, you know, quite possibly, you know, possibly just straightforward. Yeah. Is it worth saying more about your idea that, you know, AI ran out is really in single player mode? And there's opportunities for machine-age networks Sure. I mean, I mean, just a simple idea, right, is that, you know, in the Internet era, most most technologies have had sort of multiple phases where you sort of first people will build some like you have the initial early iPhone apps or like these little flashlights and other kinds of sort of simple single single player apps, something you use on your phone. is sort of a commerce network. You had, you know, Instagram, which is a social network and so forth. So one observation one might make when you're looking at, the AI landscape is amazing as it all is, it's still mostly you're sitting in front of a computer and you're typing and getting responses. There's some like little minor light sharing features. But at some point, one might expect to have, you know, sort of a network, a multiplayer phase of AI where, you know, you see it also like in a hugging face hack, there's a swarm. That's obviously a bad negative use case of a swarm. Are there positive use case? You know, you also saw I think a swarm is one of the things that solved one of the math theorems. There's, you know, another word for swarm is a, you know, collaborative fleet of agents or something. Right. And so, like, you can imagine lots of use cases where solving a hard problem, designing a building, you know, doing some scientific breakthrough. You might might that might be the right way to do it. You also see like I think Muse and Instinct have these features where you can like trust other people and you can imagine that sort of evolves into a trust network. So if you kind of squint, you're seeing the early signs of this. And I think people are, you know, just kind of an experimental phase. It's just – based on, you know, history and what might, you know, the kind of the possibilities of these products, it seems like it might be a fertile area for entrepreneurs. But by the way, we've alluded, we talked a bit about sort of the education and just, you know, the learning opportunities that are enabled via this technology. I'm curious, you know, obviously we just, you know, uh, incubated Harwoods Andreessen Academy. Thank you for support on it. If you were 18 today, how would you think about what you might study? Obviously, even, you know, academic philosophy background. Um, I think I'm stealing Mark's line, but I think you think there's one thing to do, which is study modern alchemy, right? Yeah, yeah, yeah, yeah. Go for it. Go for it. Well, humans figure is Mark's framing, but I mean, humans just figured out how to turn sand into intelligence. And the way to learn that, I think, is math, computer science, maybe economics. I don't know. And it's, you know, like, I think it's also like, like, I think Ben and Mark were lucky enough to have our, you know, a lot of our careers. uh, happened during the growth of the internet. And I think that was, I feel very fortunate that I was able to do that. You know, I think that was, uh, something bequeathed to me by, by people before me. And that was, I'm grateful to them. Um, And, you know, I think now, look, it's obvious that, you know, that the AI stuff we're talking about now is going to be the dominant narrative of the next 20, 30 years. So I think being it doesn't have to be not to necessarily be solely technical, but I think being in some way in that in the mix and at the cutting edge. [55:00] and sort of using these tools feels like kind of the obvious thing to do. I don't know what you... I think a lot of other fields are going to be revolutionized, right? So, I mean, just, you know, the field of the classics is being revolutionized because now you've got the Vesuvius Scroll Project, and you're, like, able to use, like, AI to, like, decipher, like, dark scrolls. And, you know, there may be entire new schools of philosophy that get, you know, discovered to study classics, you know, the things, you know, lost books. Economics, I just have lunch with an economics professor, and, like, the whole field of economics is going to be transformed around this. Biology. Yeah, exactly. Direct discovery, design, by the way, design, you know, we talk about movies, but also just broadly, broadly design, creativity of all kinds, all the, all the, all the forms of art, music, music. Zoology, you're going to know what the animals are saying. This is super interesting. Like we are plausibly heading towards a world in which we're going to understand what the animals are saying, right? Animal psychology will be actually real. Yeah, yeah. So like, okay, like dog, you know, I don't know, like dog, you know, dog training is going to take on a whole new thing when they don't actually understand what the, fuck the dog is sick right i think there's an x-prize for that yeah it's like a it's like a And they're definitely – I mean anybody who has a dog, no, they talk. They're in some weird language, but they definitely are saying something. And certainly cats do. I wait the entrepreneur listening to this who comes in with a pitch. Totally, totally. And then, yeah, the other way I think about it in terms of like – yeah, thinking about like 18-year-olds or kids is – Douglas Adams had like the defining description of – the guy who wrote Hitchhiker's Guide to the Galaxy of like how technology. received on a generational level. And his famous framework was there's any new technology is received three different ways by three different age cohorts. If you're below the age of 15, it's just the natural way the world works. And I've told in other settings a story of introducing my kid to JetGPT at age seven. And I'm like, you know, this is such a big deal because the computer answers questions. And he's like, duh, like what good is a computer that doesn't answer questions, right? So it's just normal. If you're between the ages of 15 to 35, it's new, career out of it, in which case you want to learn all the new things. And if you're above the age of 35, it's unholy and evil and will surely overturn all existing structures and destroy society. I'll never answer the telephone. Exactly. Exactly. Old Manielic cloud is a recurring phenomenon. And so I think 18-year-olds right now, I mean, it's got to talk about prime time to grab these new capabilities and run with them. And, you know, it's going to be creation of entirely new fields And then, look, I mean, I think kids younger than that, I'm seeing with my kid, but like, yeah, they're just, it's just, of course you have this capability. Like, of course you do. It's just like completely obvious. Well, and actually the schools, the schools are doing us a favor. Like, so the schools are not responding well to this. And so, you know, the schools are, you know, banning AI or having all these, you know, controversies with it. But of course, if you're a kid, that just makes it cooler. Yeah. Right. And so AI is going to be like the, you know, AI is your like secret superpower that you have to like deviously figure out how to take advantage of. And so I think kids are going to have all kinds of ideas for what they're going to do. adults keep saying, oh, well, what should the kids do? Like, it's so hard to be a kid. It's like, No, no, no, no. basically, It's you that's got the problem because you're obsolete. The kids are going to be all superhumans. And they're going to be able to do anything and the world is open to them and they're going to, you know, they're going to be able to enter fields that they could never enter before. What I see in this conversation and your fund and your work more generally is, to use another tool, is kind of definite optimism. The future is happening. It's going to happen. You know, a lot of what we talk about in science fiction is going to happen. It's a matter of when. And so we're going to bet early and maybe sometimes be too early, but we're going to keep betting. It's going to work out. Yeah. And to be clear, it's one thing I want to clarify is that, you know, we're throwing out all these theories and everything. But in the end, I think, as Ben said in the beginning, like we're we're in the betting on great entrepreneurs business. And so, you know, we've plenty of ideas that were not discussed today that are probably, you know, more interesting and more creative. willing to engage with those kinds of creative ideas. So I would say, you said definite optimism, I'm like optimism, but also, I would say, humble, definite optimism or something, intellectually humble. Yeah, that I've been wrong so many times. We just we really want to we want to find great people who have their own definite optimism and back them. The entrepreneurs get to be yeah. Yeah. Yep. That's a wrap. Chris, thank you guys for coming on the podcast. Good. Thank you. All right. Okay, thank you. Thanks for listening to this episode of the A16C. [60:00] podcast. If you liked this episode, be sure to like, comment, subscribe, leave us a rating or review and share it with your friends and family. For more episodes, go to YouTube, Apple Podcasts and Spotify. Follow us on X at A16Z and subscribe to our sub stack at A16Z.substack.com. Thanks again for listening and I'll see you in the next episode. As a reminder, the content here is for informational purposes only. It 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. Please note that A16Z and its affiliates may also maintain investments in the companies discussed in this podcast. For more details, including a link to our investments, please see a16z.com forward slash disclosures. Thank you.