章节 1：RFS 2026春季刊导读

📝 本节摘要：

本章作为开篇，重申了 Y Combinator 发布“创业需求”（RFS）的传统，旨在为创始人提供灵感而非强制性选题。YC 强调，随着 AI 原生技术的普及，初创公司的构建效率与野心达到了前所未有的高度。本期关注点涵盖 AI 工作流、新金融原语及工业现代化，并首次引入了来自 YC 社区创始人的前沿观察。

[原文] Requests for Startups RFS is our tradition of sharing ideas we'd like to see founders tackle.

[译文] “创业需求”（RFS）是我们的传统，旨在分享我们希望看到创始人去解决的想法。

[原文] These represent just a fraction of what we fund — if one excites you, take it as extra validation to dive in, but you don't need to work on these ideas to apply to YC.

[译文] 这些仅代表我们资助项目的一小部分——如果其中某个想法让你感到兴奋，请将其视为投身其中的额外验证，但你无需从事这些想法也能申请 YC。

[原文] Spring 2026

[译文] 2026 春季刊

[原文] The way startups are built has shifted quickly.

[译文] 初创公司的构建方式已经发生了快速转变。

[原文] AI-native companies can now be built faster, cheaper, and with more ambition than ever.

[译文] 现在的 AI 原生公司可以比以往任何时候都更快、更低成本且更具野心地构建。

[原文] We're excited about a range of startup ideas that span AI-native workflows, new financial primitives, modernized industrial systems, and more.

[译文] 我们对一系列创业想法感到兴奋，这些想法涵盖了 AI 原生工作流、新的金融原语、现代化的工业系统等等。

[原文] This time around, a few even come directly from YC founders sharing opportunities they're seeing on the frontier.

[译文] 这一次，甚至有一些想法直接来自 YC 的创始人，分享他们在前沿领域看到的机会。

章节 2：产品经理的“Cursor”

📝 本节摘要：

本章由 YC 合伙人 Andrew Miklas 撰写。他指出，尽管代码生成工具（如 Cursor）已大幅降低了软件构建门槛，但决定“构建什么”——即产品管理的核心工作（用户调研、需求定义）——仍缺乏系统性的 AI 支持。YC 正在寻找能够打造“产品管理版 Cursor”的创始人，这类工具应能利用 AI 综合分析数据，生成功能大纲及任务分解，从而连接产品发现与 AI 编码执行的完整闭环。

[原文] Cursor for Product Managers

[译文] 产品经理的“Cursor”

[原文] By Andrew Miklas

[译文] 作者：Andrew Miklas

[原文] [Andrew Miklas]: Over the last few years, we've seen an explosion of AI tools for writing code.

[译文] [Andrew Miklas]: 在过去几年里，我们看到了用于编写代码的 AI 工具呈爆炸式增长。

[原文] [Andrew Miklas]: Cursor and Claude Code are great at helping teams build software once it's clear what needs to be built.

[译文] [Andrew Miklas]: 一旦明确了需要构建什么，Cursor 和 Claude Code 在帮助团队构建软件方面表现出色。

[原文] [Andrew Miklas]: But writing code is only part of building a product people want.

[译文] [Andrew Miklas]: 但编写代码只是构建人们想要的产品的一部分。

[原文] [Andrew Miklas]: The most important part is figuring out what to build in the first place!

[译文] [Andrew Miklas]: 最重要的部分是首先弄清楚要构建什么！

[原文] [Andrew Miklas]: Every successful product requires product management: talking to users, understanding markets, synthesizing feedback, and deciding what problems are worth solving and how the product should work.

[译文] [Andrew Miklas]: 每一个成功的产品都需要产品管理：与用户交谈、了解市场、综合反馈，并决定哪些问题值得解决以及产品应如何运作。

[原文] [Andrew Miklas]: Whether this process is done by founders, engineers, or product managers, the activity is the same.

[译文] [Andrew Miklas]: 无论这个过程是由创始人、工程师还是产品经理完成，其活动本质是一样的。

[原文] [Andrew Miklas]: Historically, the output has been product requirements docs, Figma mocks, and Jira tickets — artifacts designed to communicate intent to human engineers.

[译文] [Andrew Miklas]: 历史上，其产出一直是产品需求文档、Figma 原型图和 Jira 工单——这些旨在向人类工程师传达意图的产物。

[原文] [Andrew Miklas]: Today, teams use AI in isolated parts of this process, but there's no system that supports the full loop of product discovery.

[译文] [Andrew Miklas]: 今天，团队在这个过程的孤立环节中使用 AI，但尚无系统能够支持产品发现的完整闭环。

[原文] [Andrew Miklas]: Imagine a tool where you upload customer interviews and product usage data, ask "what should we build next?", and get the outline of a new feature complete with an explanation based on customer feedback as to why this is a change worth making.

[译文] [Andrew Miklas]: 想象这样一个工具：你上传客户访谈和产品使用数据，询问“我们接下来应该构建什么？”，然后获得一个新功能的大纲，其中包含基于客户反馈的解释，说明为什么这是一个值得进行的改变。

[原文] [Andrew Miklas]: The tool would also propose specific changes to your product's UI, data model, and workflows, and would break down the development tasks so they could be handled by your favorite coding agent.

[译文] [Andrew Miklas]: 该工具还会针对你产品的 UI、数据模型和工作流提出具体的修改建议，并将开发任务进行拆解，以便由你喜欢的编码 Agent 来处理。

[原文] [Andrew Miklas]: We think there's an opportunity to build a "Cursor for product management": an AI-native system focused on helping teams figure out what to build, not just how to build it.

[译文] [Andrew Miklas]: 我们认为有机会构建一个“产品管理的 Cursor”：一个专注于帮助团队弄清楚构建什么，而不仅仅是如何构建的 AI 原生系统。

[原文] [Andrew Miklas]: As agents increasingly take the first pass at implementation, the way we define and communicate "what to build" needs to change.

[译文] [Andrew Miklas]: 随着 Agent 越来越多地承担实施的初步工作，我们需要改变定义和传达“构建什么”的方式。

[原文] [Andrew Miklas]: If you're building in this space, we'd love to hear from you.

[译文] [Andrew Miklas]: 如果你正在这个领域进行构建，我们很希望能收到你的消息。

章节 3：AI原生对冲基金

📝 本节摘要：

本章由曾任量化研究员的 Charlie Holtz 撰写。他将当前的 AI 浪潮比作 20 世纪 80 年代量化交易的崛起，指出传统大型基金在采纳新技术（如 ChatGPT）方面反应迟钝。Holtz 认为，未来的对冲基金不会仅仅将 AI 作为现有策略的附加品，而是会利用“AI Agent 集群”自主分析财报（10-Ks）、电话会议及 SEC 文件，从而发掘全新的超额收益（Alpha）来源。

[原文] AI-Native Hedge Funds

[译文] AI 原生对冲基金

[原文] By Charlie Holtz

[译文] 作者：Charlie Holtz

[原文] [Charlie Holtz]: In the 1980s, a small group of funds started using computers to analyze markets.

[译文] [Charlie Holtz]: 20 世纪 80 年代，一小群基金开始使用计算机分析市场。

[原文] [Charlie Holtz]: At the time it seemed silly, but quantitative trading is now obvious.

[译文] [Charlie Holtz]: 当时这看起来很傻，但量化交易现在已是显而易见的主流。

[原文] [Charlie Holtz]: We're at a similar inflection point now, and the next Renaissance, Bridgewater, and D.E. Shaw's are going to be built on AI.

[译文] [Charlie Holtz]: 我们现在正处于类似的拐点，下一个文艺复兴科技（Renaissance）、桥水（Bridgewater）和 D.E. Shaw 将建立在 AI 之上。

[原文] [Charlie Holtz]: The biggest funds in the world have been slow to adapt.

[译文] [Charlie Holtz]: 世界上最大的基金适应得很慢。

[原文] [Charlie Holtz]: I worked as a quant researcher at one of these funds, and when I asked compliance to let us use ChatGPT, I didn't even get a response.

[译文] [Charlie Holtz]: 我曾在其中一家基金担任量化研究员，当我请求合规部门允许我们使用 ChatGPT 时，甚至没有收到回复。

[原文] [Charlie Holtz]: It made it clear to me that the hedge funds of the future won't just bolt AI onto their existing strategies.

[译文] [Charlie Holtz]: 这让我清楚地意识到，未来的对冲基金不会只是将 AI 简单地拼接到其现有策略上。

[原文] [Charlie Holtz]: They'll use it to come up with entirely new ones. That's where the alpha is.

[译文] [Charlie Holtz]: 它们将利用 AI 来构想全新的策略。这才是超额收益（Alpha）的所在。

[原文] [Charlie Holtz]: We've already got swarms of Claude agents writing our codebases.

[译文] [Charlie Holtz]: 我们已经有成群的 Claude Agent 在编写我们的代码库。

[原文] [Charlie Holtz]: Imagine swarms of agents doing what hedge fund traders do now - combing through 10-Ks, earnings calls, and SEC filings, synthesizing analyst ideas and making trades.

[译文] [Charlie Holtz]: 想象一下成群的 Agent 做着现在对冲基金交易员所做的事情——梳理 10-K 表格、财报电话会议和 SEC 文件，综合分析师的观点并进行交易。

[原文] [Charlie Holtz]: An AI-native hedge fund will be the first to do this well.

[译文] [Charlie Holtz]: 一家 AI 原生的对冲基金将是第一个做好这件事的机构。

章节 4：AI原生代理商

📝 本节摘要：

本章由 YC 合伙人 Aaron Epstein 撰写。他分析了传统代理商模式（Agencies）难以规模化的痛点（低利润、依赖人力堆砌）。Epstein 提出，AI 将彻底改变这一局面：未来的代理商不再是“卖软件给客户用”，而是“自己用软件生成成品卖给客户”，从而获得软件公司级别的利润率。例如，设计公司可在签约前用 AI 生成定制方案，律所可用 AI 分钟级生成法律文档。

[原文] AI-Native Agencies

[译文] AI 原生代理商

[原文] By Aaron Epstein

[译文] 作者：Aaron Epstein

[原文] [Aaron Epstein]: Agencies have always been crazy hard to scale. Low margins, slow manual work, and the only way to grow is to add more people.

[译文] [Aaron Epstein]: 代理商一直都极难规模化。利润微薄，手工工作缓慢，而且唯一的增长方式就是增加更多的人手。

[原文] [Aaron Epstein]: But AI changes this.

[译文] [Aaron Epstein]: 但 AI 改变了这一点。

[原文] [Aaron Epstein]: Now instead of selling software to customers to help them do the work, you can charge way more by using the software yourself and selling them the finished product at 100x the price.

[译文] [Aaron Epstein]: 现在，你不再是向客户出售软件来帮助他们工作，而是可以自己使用软件，并将成品以 100 倍的价格卖给他们，从而收取高得多的费用。

[原文] [Aaron Epstein]: Think of a design firm that uses AI to produce custom design work for clients upfront, to win the business before the contract is even signed.

[译文] [Aaron Epstein]: 想象一家设计公司，在合同签署之前就利用 AI 预先为客户制作定制设计作品，以此赢得业务。

[原文] [Aaron Epstein]: Or an ad agency that uses AI to create stunning video ads without the time and expense of setting up a physical shoot.

[译文] [Aaron Epstein]: 或者一家广告代理商，利用 AI 制作令人惊叹的视频广告，而无需花费时间和金钱去搭建实体拍摄场景。

[原文] [Aaron Epstein]: Or a law firm that uses AI to write legal docs in minutes, rather than weeks.

[译文] [Aaron Epstein]: 又或者一家律师事务所，利用 AI 在几分钟内（而不是几周）起草法律文件。

[原文] [Aaron Epstein]: That's why agencies of the future will look more like software companies, with software margins.

[译文] [Aaron Epstein]: 这就是为什么未来的代理商看起来会更像软件公司，拥有软件行业的利润率。

[原文] [Aaron Epstein]: And they'll scale far bigger than any agencies that exist in these fragmented markets today.

[译文] [Aaron Epstein]: 而且它们的规模将远超当今这些分散市场中存在的任何代理商。

[原文] [Aaron Epstein]: If you're rethinking how agencies and service businesses of the future will be built, we'd love to hear from you.

[译文] [Aaron Epstein]: 如果你正在重新思考未来的代理商和服务型企业将如何构建，我们很希望能收到你的消息。

章节 5：稳定币金融服务

📝 本节摘要：

本章由 Daivik Goel 撰写。他指出，尽管稳定币已成为全球金融的关键基础设施，但基于此的金融服务层仍有巨大空白。随着 GENIUS 和 CLARITY 法案的推出，稳定币处于连接 DeFi（去中心化金融）与 TradFi（传统金融）的独特合规位置。YC 认为现在是构建“中间地带”服务的最佳时机——即在合规框架下，提供具备 DeFi 高收益和代币化资产优势的金融产品，解决传统金融收益低与加密市场风险高之间的两难困境,,。

[原文] Stablecoin Financial Services

[译文] 稳定币金融服务

[原文] By Daivik Goel

[译文] 作者：Daivik Goel

[原文] [Daivik Goel]: Stablecoins are rapidly becoming critical infrastructure for global finance, yet much of the financial services layer remains unbuilt.

[译文] [Daivik Goel]: 稳定币正迅速成为全球金融的关键基础设施，然而其上的金融服务层大部分仍未建立。

[原文] [Daivik Goel]: The GENIUS and CLARITY Acts are placing stablecoins in a unique position between DeFi and TradFi, compliant but crypto-native.

[译文] [Daivik Goel]: GENIUS 和 CLARITY 法案将稳定币置于 DeFi（去中心化金融）和 TradFi（传统金融）之间的独特位置，既合规又具备加密原生特性。

[原文] [Daivik Goel]: This creates room for financial services that offer DeFi benefits like better yield or access to tokenized real-world assets while operating under traditional compliance frameworks.

[译文] [Daivik Goel]: 这为那些在传统合规框架下运营，同时提供 DeFi 优势（如更高收益率或访问代币化现实世界资产）的金融服务创造了空间。

[原文] [Daivik Goel]: Today, businesses and individuals must choose between regulated financial products with limited upside and unregulated crypto with real risk.

[译文] [Daivik Goel]: 如今，企业和个人必须在收益有限的受监管金融产品和具有真正风险的不受监管加密货币之间做出选择。

[原文] [Daivik Goel]: Stablecoins sitting in the regulatory middle ground can bridge this gap, whether that's yield-bearing accounts, new investment access, or infrastructure that makes money move faster and cheaper across borders.

[译文] [Daivik Goel]: 处于监管中间地带的稳定币可以弥合这一差距，无论是生息账户、新的投资渠道，还是让资金跨境流动更快、更便宜的基础设施。

[原文] [Daivik Goel]: The regulatory window is open. The rails are being laid.

[译文] [Daivik Goel]: 监管窗口已经打开。轨道正在铺设中。

[原文] [Daivik Goel]: It's the perfect time to build something that blurs the line between the two worlds.

[译文] [Daivik Goel]: 现在正是构建模糊这两个世界界限的事物的最佳时机。

章节 6：政府AI应用

📝 本节摘要：

本章由 YC 合伙人 Tom Blomfield 撰写。他敏锐地观察到一个供需不对称现象：第一波 AI 工具极大地加速了公众“提交表格”的速度，导致政府端接收到的申请量激增，但政府内部仍普遍采用手工打印处理的落后模式。Tom 呼吁创业者开发针对政府的 AI 基础设施，以应对这一即将到来的行政负荷冲击。尽管 B2G（对政府销售）难度极高，但这类客户具备极高的留存率（粘性）和巨大的合同扩展潜力。

[原文] AI for Government

[译文] 政府 AI 应用

[原文] By Tom Blomfield

[译文] 作者：Tom Blomfield

[原文] [Tom Blomfield]: The first wave of AI companies has helped businesses and normal people fill in forms and complete online applications with unprecedented speed and accuracy.

[译文] [Tom Blomfield]: 第一波 AI 公司已经帮助企业和普通人以前所未有的速度和准确性填写表格并完成在线申请。

[原文] [Tom Blomfield]: On the flip side, many of these forms will be received by local, state, and federal government, where they're currently printing them out and processing them by hand.

[译文] [Tom Blomfield]: 但反过来看，这些表格中的许多将由地方、州和联邦政府接收，而目前他们仍在将这些表格打印出来并手工处理。

[原文] [Tom Blomfield]: Government desperately needs AI tools to deal with the huge increase that's coming down the line.

[译文] [Tom Blomfield]: 政府迫切需要 AI 工具来应对即将到来的巨大增长。

[原文] [Tom Blomfield]: And the benefit is that it will also make government much more cost-effective and responsive.

[译文] [Tom Blomfield]: 其好处在于，这也将使政府更具成本效益，反应更敏捷。

[原文] [Tom Blomfield]: We've seen hints of this digital government in places like Estonia, but we need to spread it to the rest of the world.

[译文] [Tom Blomfield]: 我们已经在爱沙尼亚等地看到了这种数字政府的雏形，但我们需要将其推广到世界其他地方。

[原文] [Tom Blomfield]: This kind of startup is not for the faint of heart.

[译文] [Tom Blomfield]: 这类创业并不适合胆小的人。

[原文] [Tom Blomfield]: Selling to government is extremely hard, but once you've figured out how to land your first customer, they tend to be very sticky and can expand to huge contracts.

[译文] [Tom Blomfield]: 向政府销售极其困难，但一旦你弄清楚如何拿下第一个客户，他们往往具有很高的粘性，并且可以扩展成巨大的合同。

章节 7：现代金属工厂

📝 本节摘要：

本章由 Zane Hengsperger 撰写。他指出美国制造业复兴不仅仅关乎劳动力成本，核心痛点在于金属工厂（特别是铝材和钢管）的设计陈旧、运转缓慢且效率低下。目前的工厂优化吨位而非速度，自动化程度低且能源利用僵化。Zane 认为，利用 AI 驱动的生产计划、实时 MES 系统及现代能源管理，构建“软件定义”的新一代金属工厂，将是降低国内金属成本、缩短交货周期并重塑美国工业基础的巨大机遇。

[原文] Modern Metal Mills

[译文] 现代金属工厂

[原文] By Zane Hengsperger

[译文] 作者：Zane Hengsperger

[原文] [Zane Hengsperger]: When people talk about reindustrializing America, they usually focus on labor costs or geopolitics.

[译文] [Zane Hengsperger]: 当人们谈论美国再工业化时，他们通常关注劳动力成本或地缘政治。

[原文] [Zane Hengsperger]: But a bigger problem is hiding in plain sight: American metal mills are slow by design.

[译文] [Zane Hengsperger]: 但一个更大的问题正隐藏在显眼处：美国金属工厂的设计本质上就是缓慢的。

[原文] [Zane Hengsperger]: If you buy rolled aluminum or steel tube in the U.S., lead times of 8 to 30 weeks are normal.

[译文] [Zane Hengsperger]: 如果你在美国购买轧制铝材或钢管，8 到 30 周的交货时间是常态。

[原文] [Zane Hengsperger]: Most buyers can't even purchase directly from mills.

[译文] [Zane Hengsperger]: 大多数买家甚至无法直接从工厂购买。

[原文] [Zane Hengsperger]: And despite high prices, mills still operate on thin margins.

[译文] [Zane Hengsperger]: 尽管价格高昂，工厂的利润率却依然微薄。

[原文] [Zane Hengsperger]: That's not because demand is weak or workers are unskilled—it's because the systems running these mills were designed decades ago.

[译文] [Zane Hengsperger]: 这不是因为需求疲软或工人技能不足——而是因为运行这些工厂的系统是几十年前设计的。

[原文] [Zane Hengsperger]: Production planning, scheduling, quoting, and execution are fragmented.

[译文] [Zane Hengsperger]: 生产计划、调度、报价和执行都是碎片化的。

[原文] [Zane Hengsperger]: Mills optimize for tonnage and utilization, not speed, flexibility, or margin.

[译文] [Zane Hengsperger]: 工厂优化的是吨位和利用率，而不是速度、灵活性或利润率。

[原文] [Zane Hengsperger]: Short runs and spec changes are treated as disruptions instead of opportunities.

[译文] [Zane Hengsperger]: 短期运行和规格变更被视为干扰而非机会。

[原文] [Zane Hengsperger]: Automation has lagged at the exact moment the workforce is shrinking.

[译文] [Zane Hengsperger]: 在劳动力萎缩的时刻，自动化却滞后了。

[原文] [Zane Hengsperger]: Material handling, changeovers, inspection, and quality control still rely on tribal knowledge held by a few experienced operators.

[译文] [Zane Hengsperger]: 物料搬运、换型、检查和质量控制仍然依赖于少数经验丰富的操作员所掌握的部落知识（隐性知识）。

[原文] [Zane Hengsperger]: Automation is mostly used to push more tons through a slow system, not to eliminate setup time or variability.

[译文] [Zane Hengsperger]: 自动化主要用于推动更多吨位通过缓慢的系统，而不是消除设置时间或变异性。

[原文] [Zane Hengsperger]: Energy is the other half of the problem.

[译文] [Zane Hengsperger]: 能源是问题的另一半。

[原文] [Zane Hengsperger]: Aluminum and steel are extremely energy-intensive, yet most mills rely on legacy power contracts and inflexible grids.

[译文] [Zane Hengsperger]: 铝和钢是极其能源密集型的，然而大多数工厂依赖于旧有的电力合同和不灵活的电网。

[原文] [Zane Hengsperger]: New energy models—on-site generation, smarter power management, even next-generation nuclear—could dramatically reduce costs, but they're rarely designed into mills from the start.

[译文] [Zane Hengsperger]: 新的能源模式——现场发电、更智能的电源管理，甚至下一代核能——可以大幅降低成本，但它们很少从一开始就被设计进工厂。

[原文] [Zane Hengsperger]: What's changed is that software and energy technology are finally good enough to rethink the entire system.

[译文] [Zane Hengsperger]: 改变的是，软件和能源技术终于好到可以重新思考整个系统了。

[原文] [Zane Hengsperger]: AI-driven planning, real-time MES, and modern automation can compress lead times and raise margins at the same time.

[译文] [Zane Hengsperger]: AI 驱动的计划、实时制造执行系统（MES）和现代自动化可以同时压缩交货时间并提高利润率。

[原文] [Zane Hengsperger]: We think this creates an opportunity to build modern, software-defined American mills—especially in aluminum rolling and steel tube—where long lead times and energy costs are most entrenched.

[译文] [Zane Hengsperger]: 我们认为这创造了一个构建现代、软件定义的美国工厂的机会——特别是在交货时间和能源成本最根深蒂固的铝轧制和钢管领域。

[原文] [Zane Hengsperger]: Modernizing mills isn't just about going faster.

[译文] [Zane Hengsperger]: 现代化工厂不仅仅是关于走得更快。

[原文] [Zane Hengsperger]: It's about making domestic metal cheaper, more flexible, and more profitable—and rebuilding the industrial foundation of the U.S.

[译文] [Zane Hengsperger]: 它是关于使国内金属更便宜、更灵活、更有利可图——并重建美国的工业基础。

章节 8：体力工作AI指导

📝 本节摘要：

本章由 David Lieb 撰写。他借用《黑客帝国》中尼奥瞬间学会功夫的经典桥段，形象地描述了 AI 对实体工作的赋能前景。不同于脑机接口，David 提出利用多模态 AI 模型结合普及的硬件（手机、智能眼镜），通过实时视觉识别与语音指导，让工人在“实战”中即时获得技能。这一模式旨在解决熟练劳动力短缺问题，将复杂的维修、制造或护理工作转化为有 AI 辅助的高效任务，无论是作为工具软件出售、构建垂直领域的“超级劳动力”，还是打造零工经济平台，都蕴含巨大潜力。

[原文] AI Guidance for Physical Work

[译文] 体力工作 AI 指导

[原文] By David Lieb

[译文] 作者：David Lieb

[原文] [David Lieb]: You know that scene in The Matrix, where Neo plugs a cable into the back of his head and wakes up a while later and says "I know Kung Fu"?

[译文] [David Lieb]: 你知道《黑客帝国》里的那个场景吗，尼奥把一根缆线插到后脑勺，过了一会儿醒来说“我会功夫了”？

[原文] [David Lieb]: Physical work is about to get something similar – not through brain implants, but through real-time AI guidance.

[译文] [David Lieb]: 体力工作即将迎来类似的东西——不是通过脑部植入物，而是通过实时 AI 指导。

[原文] [David Lieb]: Much of the AI conversation focuses on which desk jobs will get replaced.

[译文] [David Lieb]: 关于 AI 的讨论大多集中在哪些案头工作将被取代。

[原文] [David Lieb]: But for physical work—stuff like field services, manufacturing, healthcare—AI can't yet act in the world.

[译文] [David Lieb]: 但对于体力工作——比如现场服务、制造业、医疗保健——AI 还无法在现实世界中行动。

[原文] [David Lieb]: What it can do is see, reason, and guide the human who does.

[译文] [David Lieb]: 它能做的是“看”、“推理”，并指导正在行动的人类。

[原文] [David Lieb]: Imagine wearing a small camera while an AI sees what you see and talks you through the job: "turn off that valve", "use the ⅜ inch wrench", "that part looks worn, replace it".

[译文] [David Lieb]: 想象一下戴着一个小摄像头，AI 看到你所看到的，并通过语音指导你完成工作：“关掉那个阀门”、“使用 ⅜ 英寸的扳手”、“那个部件看起来磨损了，更换它”。

[原文] [David Lieb]: Instead of needing months or years of training, workers can become effective immediately, with AI coaching them and accessing new skills when needed.

[译文] [David Lieb]: 工人无需数月或数年的培训，在 AI 的指导下并在需要时获取新技能，就能立即变得高效。

[原文] [David Lieb]: Why now? Three things have converged.

[译文] [David Lieb]: 为什么是现在？三件事汇聚在了一起。

[原文] [David Lieb]: First, multimodal models can now see and reason about real-world situations reliably.

[译文] [David Lieb]: 首先，多模态模型现在可以可靠地“看”并推理现实世界的情况。

[原文] [David Lieb]: Second, the hardware is already everywhere – phones, AirPods, Smart Glasses.

[译文] [David Lieb]: 其次，硬件已经无处不在——手机、AirPods、智能眼镜。

[原文] [David Lieb]: And third, skilled labor shortages make this economically urgent and a high wage job for millions of people.

[译文] [David Lieb]: 第三，熟练劳动力的短缺使得这在经济上变得紧迫，并为数百万人提供了高薪工作。

[原文] [David Lieb]: There are a few approaches you could take.

[译文] [David Lieb]: 你可以采取几种方法。

[原文] [David Lieb]: The most obvious is to build this system and sell it to companies with existing workforces.

[译文] [David Lieb]: 最明显的是构建这个系统并将其出售给拥有现有劳动力的公司。

[原文] [David Lieb]: Or, you could pick a vertical, like HVAC repair or nursing, and build a full-stack superpowered workforce.

[译文] [David Lieb]: 或者，你可以选择一个垂直领域，比如暖通空调（HVAC）维修或护理，并建立一支全栈式的超能力劳动力队伍。

[原文] [David Lieb]: Or, you could build a platform that lets anyone sign up and become a skilled worker or start their own business.

[译文] [David Lieb]: 或者，你可以建立一个平台，让任何人都能注册成为熟练工人或开始他们自己的业务。

[原文] [David Lieb]: If you're interested in giving physical workers the same type of AI superpowers that Claude Code gives you, we'd love to see you apply.

[译文] [David Lieb]: 如果你有兴趣赋予体力劳动者像 Claude Code 赋予你那样的 AI 超能力，我们很希望看到你的申请。

章节 9：大型空间模型

📝 本节摘要：

本章由 Ryan McLinko 撰写。他认为尽管大语言模型（LLM）推动了 AI 的近期突破，但其局限在于仅能处理语言表达的领域。要解锁 AGI（通用人工智能），必须开发具备“空间推理”能力的新一代模型。Ryan 指出，当前的 AI 难以处理复杂的 2D/3D 空间操作（如心理旋转），未来的机会在于构建将几何与物理结构作为“一等公民”的基础模型，这类公司的潜力足以比肩 OpenAI 或 Anthropic。

[原文] Large Spatial Models

[译文] 大型空间模型

[原文] By Ryan McLinko

[译文] 作者：Ryan McLinko

[原文] [Ryan McLinko]: Large language models have driven most of the recent breakthroughs in AI, but their impact has been constrained to domains that can be expressed primarily through language.

[译文] [Ryan McLinko]: 大语言模型推动了 AI 近期的大部分突破，但其影响仅局限于那些主要可以通过语言表达的领域。

[原文] [Ryan McLinko]: Unlocking the next wave of AI capability, and enabling artificial general intelligence, will require models that are capable of spatial reasoning.

[译文] [Ryan McLinko]: 解锁下一波 AI 能力并实现通用人工智能，将需要具备空间推理能力的模型。

[原文] [Ryan McLinko]: Today's systems can handle limited spatial tasks, such as basic relationships or depth estimation, but they cannot robustly reason about spatial manipulation, 2D and 3D features, their relationships, or operations like mental rotation.

[译文] [Ryan McLinko]: 今天的系统可以处理有限的空间任务，例如基本关系或深度估计，但它们无法稳健地推理空间操作、2D 和 3D 特征、它们之间的关系或像心理旋转这样的运算。

[原文] [Ryan McLinko]: This limits AI's ability to understand and interact with the physical world.

[译文] [Ryan McLinko]: 这限制了 AI 理解物理世界并与之交互的能力。

[原文] [Ryan McLinko]: There is an opportunity to build large-scale spatial reasoning models that treat geometry and physical structure as first-class primitives, not approximations layered on top of language.

[译文] [Ryan McLinko]: 有机会构建大规模的空间推理模型，将几何和物理结构视为一等原语（First-class primitives），而不是叠加在语言之上的近似值。

[原文] [Ryan McLinko]: Such models would enable AI systems to reason about and design real-world objects and environments.

[译文] [Ryan McLinko]: 这样的模型将使 AI 系统能够推理和设计现实世界的物体及环境。

[原文] [Ryan McLinko]: A company that succeeds in building this capability could define the next AI foundation model, on the scale of OpenAI or Anthropic.

[译文] [Ryan McLinko]: 成功构建这种能力的公司可能会定义下一个 AI 基础模型，达到 OpenAI 或 Anthropic 的规模。

章节 10：政府反欺诈基础设施

📝 本节摘要：

本章由 YC 总裁 Garry Tan 撰写。他聚焦于政府作为全球最大客户面临的巨额资金流失问题（仅 Medicare 每年损失数百亿）。Garry 提出了利用《虚假申报法》中的“公益代位诉讼”（Qui Tam）条款进行资金追回的商业模式。目前该流程极度依赖律所的人工操作，效率低下。YC 呼吁具备法律、合规或审计背景的创始人，利用 AI 构建智能基础设施，自动解析杂乱的证据文件并生成诉状，从而将反欺诈流程加速 10 倍，为纳税人挽回数十亿美元。

[原文] Infra for Government Fraud Hunters

[译文] 政府反欺诈基础设施

[原文] By Garry Tan

[译文] 作者：Garry Tan

[原文] [Garry Tan]: We want to fund startups that bring government fraud investigation into the modern era.

[译文] [Garry Tan]: 我们希望资助那些将政府欺诈调查带入现代时代的初创公司。

[原文] [Garry Tan]: Government is the biggest customer on earth—it spends trillions annually at the federal, state and local levels, and it hemorrhages a commensurate amount in fraud.

[译文] [Garry Tan]: 政府是地球上最大的客户——它每年在联邦、州和地方各级花费数万亿美元，并因欺诈而流失相应数量的巨额资金。

[原文] [Garry Tan]: Medicare alone loses tens of billions a year to improper payments.

[译文] [Garry Tan]: 仅医疗保险（Medicare）一项，每年因不当支付造成的损失就达数百亿美元。

[原文] [Garry Tan]: One of the most effective ways to claw this money back at scale is the qui tam provision under the False Claims Act.

[译文] [Garry Tan]: 大规模追回这笔钱的最有效方法之一是《虚假申报法》（False Claims Act）下的“公益代位诉讼”（qui tam）条款。

[原文] [Garry Tan]: This lets private citizens file lawsuits on behalf of the government against companies defrauding it.

[译文] [Garry Tan]: 这允许普通公民代表政府起诉那些欺诈政府的公司。

[原文] [Garry Tan]: If the case succeeds, these citizens get to keep a percentage of whatever's recovered.

[译文] [Garry Tan]: 如果案件胜诉，这些公民可以保留追回款项的一定比例。

[原文] [Garry Tan]: At the moment, this process is painfully slow: An insider tips off a law firm, and then the firm spends months or years manually pulling documents and building the case.

[译文] [Garry Tan]: 目前，这个过程极其缓慢：知情人士向律师事务所举报，然后律所花费数月或数年时间手动提取文件并建立案件。

[原文] [Garry Tan]: This should be accelerated with software.

[译文] [Garry Tan]: 这应该通过软件来加速。

[原文] [Garry Tan]: Not dashboards, but intelligent systems that can take an insider tip and organize the evidence around it—parsing messy PDFs, tracing opaque corporate structures, and packaging the findings into complaint-ready files.

[译文] [Garry Tan]: 不是仪表盘，而是智能系统，它能获取内部线索并围绕其组织证据——解析混乱的 PDF，追踪不透明的公司结构，并将调查结果打包成可直接用于投诉的文件。

[原文] [Garry Tan]: Some startups are already filing FCA claims themselves, but we think there's a big opportunity to build tools that dramatically speed up whistleblower law firms, state AGs, and inspectors general.

[译文] [Garry Tan]: 一些初创公司已经开始自己提起 FCA（虚假申报法）索赔，但我们认为存在一个巨大的机会，可以构建工具来大幅加快举报人律所、州总检察长（State AGs）和监察长（Inspectors General）的工作速度。

[原文] [Garry Tan]: Founder profile matters here.

[译文] [Garry Tan]: 在这里，创始人的背景很重要。

[原文] [Garry Tan]: We're looking for teams where at least one founder has actually done work like this, whether that's a former FCA counsel, compliance lead or auditor.

[译文] [Garry Tan]: 我们正在寻找这样的团队：至少有一位创始人实际从事过此类工作，无论是前 FCA 顾问、合规主管还是审计师。

[原文] [Garry Tan]: Now is the time to build this: the AI capabilities are finally here, and there's bipartisan tailwind to act.

[译文] [Garry Tan]: 现在正是构建它的时候：AI 能力终于具备了，而且有两党支持的顺风来采取行动。

[原文] [Garry Tan]: If you can make fraud recovery 10x faster, you'll build a huge business — and return billions to taxpayers.

[译文] [Garry Tan]: 如果你能使欺诈追回速度提高 10 倍，你将建立一个巨大的企业——并向纳税人返还数十亿美元。

章节 11：简化大模型训练

📝 本节摘要：

本章（最终章）由 YC 创始人 Gabriel Birnbaum 撰写。他以亲身经历指出，尽管 AI 备受瞩目，但大模型（LLM）的训练工具链依然由于 SDK 损坏、GPU 实例故障和开源工具的 Bug 而显得极其原始和困难。Gabriel 呼吁构建新一代的开发者工具——包括抽象训练过程的 API、管理海量数据集的数据库以及专为 ML 研发设计的开发环境。他预言，随着后训练（Post-training）和模型专业化日益重要，这些工具将成为未来软件构建的基础设施。

[原文] Make LLMs Easy to Train

[译文] 让 LLM 训练变得简单

[原文] By Gabriel Birnbaum

[译文] 作者：Gabriel Birnbaum

[原文] [Gabriel Birnbaum]: Training large language models is still surprisingly difficult.

[译文] [Gabriel Birnbaum]: 训练大型语言模型仍然出奇地困难。

[原文] [Gabriel Birnbaum]: My co-founder Eric and I have spent the last three years training diffusion and language models at Can of Soup, and despite all the attention AI has received, the tooling has barely improved.

[译文] [Gabriel Birnbaum]: 我的联合创始人 Eric 和我在 Can of Soup 花了过去三年时间训练扩散模型和语言模型，尽管 AI 受到了如此多的关注，但工具却几乎没有改进。

[原文] [Gabriel Birnbaum]: On any given day we may spend significant time dealing with broken SDKs, SSHing into busted GPU instances (that you only find out are busted after spinning them up for half an hour), or discovering major bugs in open-source tooling.

[译文] [Gabriel Birnbaum]: 在任何一天，我们要么花费大量时间处理损坏的 SDK，要么通过 SSH 连接到坏掉的 GPU 实例（你只有在启动半小时后才发现它们是坏的），要么发现开源工具中的重大 Bug。

[原文] [Gabriel Birnbaum]: Not to mention the work of managing, sourcing, processing, and visualizing terabytes of data.

[译文] [Gabriel Birnbaum]: 更不用说管理、寻源、处理和可视化数太字节（TB）数据的工作了。

[原文] [Gabriel Birnbaum]: I'd love to use products that make LLM training easy.

[译文] [Gabriel Birnbaum]: 我很希望能使用那些让 LLM 训练变得简单的产品。

[原文] [Gabriel Birnbaum]: • APIs that abstract training.

[译文] [Gabriel Birnbaum]: • 抽象化训练过程的 API。

[原文] [Gabriel Birnbaum]: • Databases to easily manage very large datasets.

[译文] [Gabriel Birnbaum]: • 能轻松管理超大数据集的数据库。

[原文] [Gabriel Birnbaum]: • Dev environments built with ML research in mind.

[译文] [Gabriel Birnbaum]: • 专为机器学习研究而构建的开发环境。

[原文] [Gabriel Birnbaum]: As post-training and model specialization become more important, I could see these products becoming the foundation of how software is built in the future.

[译文] [Gabriel Birnbaum]: 随着后训练和模型专业化变得越来越重要，我可以预见这些产品将成为未来软件构建的基础。