蒸汽、钢铁与无限心智：当AI超越"副驾驶"
Notion CEO Ivan Zhao以钢铁与蒸汽为隐喻，重新审视AI对知识工作、组织架构与全球经济的深层变革——我们仍在...
原文标题：Steam, Steel, and Infinite Minds
作者：Ivan Zhao（Notion CEO）
来源：Notion Blog
Every era is shaped by its miracle material. Steel forged the Gilded Age. Semiconductors switched on the Digital Age. Now AI has arrived as infinite minds. If history teaches us anything, those who master the material define the era.
每个时代都由其奇迹材料所塑造。钢铁锻造了镀金时代，半导体开启了数字时代。如今人工智能以无限心智的姿态降临。如果说历史教会了我们什么，那就是掌握核心材料者定义时代。
In the 1850s, Andrew Carnegie ran through muddy Pittsburgh streets as a telegraph boy. Six in ten Americans were farmers. Within two generations, Carnegie and his peers remade the world: horses gave way to railroads, candlelight to electricity, iron to steel.
19世纪50年代，安德鲁·卡内基还是个在匹兹堡泥泞街道上奔跑的电报童。那时，六成美国人以务农为生。仅仅两代人的时间，卡内基和他的同辈便重塑了世界：马车让位于铁路，烛火让位于电灯，生铁让位于钢铁。
Since then, work shifted from factories to offices. Today I run a software company in San Francisco, building tools for millions of knowledge workers. In this industry town, everyone talks about AGI, but most of the two billion office workers worldwide haven’t felt its impact. What will knowledge work look like soon? What happens when organizations start incorporating minds that never sleep?
自那以后，工作从工厂转移到了办公室。如今，我在旧金山经营一家软件公司，为数百万知识工作者打造工具。在这座科技重镇，人人都在谈论通用人工智能，但全球二十亿办公室工作者中的大多数，尚未真正感受到它的存在。知识工作很快会变成什么样？当组织架构开始吸纳那些永不休眠的”思维”时，会发生什么？
The future is often unrecognizable because it arrives disguised as the past. Early telephones were seen as faster telegraphs; early films as recorded plays. As Marshall McLuhan put it, “We drive into the future using only our rearview mirror.” We’re still in that uncomfortable transition phase—retrofitting AI into old waterwheels designed for humans, expecting an industrial revolution.
未来往往难以预测，因为它总是伪装成过去的模样。早期电话被视为更快捷的电报，早期电影不过是被拍摄下来的戏剧。正如马歇尔·麦克卢汉所说：”我们总是透过后视镜驶向未来。”我们仍处于那个令人不安的过渡阶段——将人工智能硬塞进为人设计的旧水车，却期待它引发一场工业革命。
Today, we see this as AI chatbots which mimic Google search boxes. We’re now deep in that uncomfortable transition phase which happens with every new technology shift. For what comes next, I don’t have all the answers. But I like to use a few historical metaphors to think about how AI might work at different scales—individuals, organizations, and entire economies.
如今，我们看到的人工智能聊天机器人，也在模仿谷歌搜索框的形态。我们正深陷于每次技术变革都会经历的那个令人不适的过渡期。对于未来，我并非无所不知，但我喜欢借助几个历史隐喻，思考人工智能如何在个人、组织乃至整个经济体等不同层面发挥作用。
Individuals: From Bicycles to Cars
个人：从自行车到汽车
The first signs of change are appearing among the priestly class of knowledge work—programmers.
变革的最初迹象，出现在知识工作的精英群体——程序员身上。
My co-founder Simon was what we call a 10x engineer. Today he rarely writes code himself. Walk past his desk, and you’ll see him directing three or four AI coding agents at once. These agents don’t just type faster—they can think, turning him into a 30x or 40x engineer. He assigns tasks over lunch or before bed, and the agents keep working while he’s away. He’s become a manager of infinite minds.
我的联合创始人西蒙曾是我们口中的”10倍程序员”，但如今他很少亲自写代码了。走过他的办公桌，你会看到他同时调度三四个人工智能编程智能体。这些智能体不只是打字更快，还具备思考能力，这让他的效率直接提升到了普通工程师的30-40倍。他会在午餐或睡前下达任务，让智能体在他离开时继续工作。他已然成为了”无限大脑的管理者”。
In the 1980s, Steve Jobs called the personal computer a “bicycle for the mind.” A decade later, we paved the “information superhighway”—the internet. But to this day, most knowledge work still runs on human power, like pedaling a bicycle on the highway for decades.
1980年，史蒂夫·乔布斯将个人电脑称作”大脑的自行车”。十年后，我们铺设了”信息高速公路”——互联网。但时至今日，大多数知识工作仍依赖人力驱动，就像我们在高速公路上蹬了几十年自行车。
With AI agents, people like Simon have upgraded from bicycles to cars. When will other types of knowledge workers get behind the wheel? Two big challenges stand in the way.
借助人工智能智能体，西蒙这样的人已经从骑自行车升级为开汽车。其他类型的知识工作者何时才能开上汽车？有两大难题横亘在前。
Why is AI less effective at general knowledge work than coding? Because it suffers from context fragmentation and lack of verifiability.
与编程智能体相比，人工智能为何更难助力通用知识工作？因为后者存在场景碎片化与结果难验证的特性。
The first missing ingredient is fragmented context. For coding, tools and context live in one place—the IDE, codebase, terminal. But general knowledge work is scattered across dozens of tools. Imagine an AI agent trying to draft a product brief. It needs to pull from Slack conversations, a strategy doc, last quarter’s metrics from a dashboard, and institutional memory that lives only in someone’s head.
第一个缺失的要素是碎片化的上下文。对于编码来说，工具和上下文往往存在于一个地方——集成开发环境、代码库、终端。但通用的知识工作分散在几十种工具中。想象一个AI智能体试图草拟一份产品简报，它需要从Slack对话、一份战略文档、仪表板中上季度的指标，以及只存在于某人脑海中的机构记忆中提取信息。
Today humans are the glue, stitching this all together via copy-paste and tab-switching. Until these contexts are integrated, agents will remain trapped in narrow use cases. The second missing ingredient is verifiability. Code has a magical property: you can verify it with tests and errors. Model builders use this to train AI to code better (e.g., reinforcement learning).
如今，人类是胶水，通过复制粘贴和在浏览器标签页之间切换将这一切缝合在一起。在这些上下文被整合之前，智能体将一直受困于狭窄的使用场景。第二个缺失的要素是可验证性。代码有个神奇属性：可以用测试和报错来验证。模型构建者利用这一点，训练AI把编程做得更好（例如强化学习）。
But how do you verify if a project is well-managed, or a strategy memo is good? We haven’t yet found ways to improve models for general knowledge work. So humans still need to be in the loop to supervise, guide, and show what good looks like.
但你如何验证一个项目是否管理得当，或一份战略备忘录是否写得好？我们还没有找到方法来改进模型，让它们胜任普遍性的知识工作。因此，人类仍需要”介入”来监督、引导，并示范”什么是好”。
The 1865 Red Flag Act required someone to walk in front of vehicles waving a red flag (it was repealed in 1896). This is an unpopular example of human-in-the-loop. This year’s experiments with coding agents have taught us that human-in-the-loop isn’t always ideal. It’s like having someone check every screw on an assembly line, or requiring a flag bearer to walk in front of a car.
1865年的《红旗法案》要求车辆行驶时须有人持红旗步行引导（该法案于1896年废止）。这是一个不受欢迎的”人在回路”的例子。今年编程智能体的实践教会我们，”人在回路”并非总是理想的。这如同安排专人检查生产线上的每一个螺丝，或要求红旗手走在汽车前开道。
We want humans to supervise from a position of leverage, not get stuck in the loop. Once context is integrated and work is verifiable, billions of workers will upgrade from bicycles to cars—and eventually to self-driving.
我们希望人类从杠杆支点进行监督，而不是身陷回路之中。一旦上下文实现整合且工作可被验证，数十亿工作者将从蹬自行车升级为驾驶汽车，最终迈向自动驾驶。
Organizations: Steel and Steam
组织：钢铁与蒸汽
Companies are a recent invention. As they grow, they suffer from diminishing returns and hit limits. Hundreds of years ago, most companies were small workshops of a dozen people. Today we have multinational corporations with hundreds of thousands of employees.
公司是近代的发明。随着规模扩大，它们会效能衰减并触及极限。几百年前，大多数公司只是十几人的作坊。如今我们拥有雇员数十万的跨国企业。
Communication infrastructure—human brains connected via meetings and messages—buckles under exponential load. We’ve tried to solve this with hierarchy, process, and documentation—but we’ve been using human-scale tools to solve industrial-scale problems, like building skyscrapers with wood.
沟通基础设施（通过会议和信息连接的人类大脑）在指数级的负荷下不堪重负。我们试图用层级、流程和文档来解决这一困局。但我们一直是在用人力尺度的工具解决工业级的问题，犹如用木材建造摩天大楼。
Two historical metaphors reveal how the new miracle material might reshape the future of organizations. The first is steel. Before steel, buildings in the 19th century had a limit of six or seven floors. Iron was strong but brittle and heavy; add more floors, and the structure collapsed under its own weight.
两个历史隐喻，揭示了在新的奇迹材料加持下，未来组织可能呈现的全新面貌。第一个是钢铁。19世纪，在钢铁出现之前，建筑的高度上限只有六七层。铁虽坚固，却又脆又重，层数再多，建筑就会因自身重量而坍塌。
Steel changed everything. It’s strong yet malleable. Frames could be lighter, walls thinner, and suddenly buildings could rise dozens of stories. New kinds of buildings became possible.
钢铁改变了一切，它坚固且具韧性，框架可以更轻，墙体可以更薄，数十层的高楼拔地而起，新的建筑形态成为可能。
AI is steel for organizations. It has the potential to maintain context across workflows and surface decisions when needed without the noise. Human communication no longer has to be the load-bearing wall. The weekly two-hour alignment meeting becomes a five-minute async review.
人工智能就是组织的钢铁。它有能力贯通工作流程，在需要时精准呈现决策，摒弃冗余噪音。人际沟通不再是组织的承重墙。原本两小时的周度对齐会议，变成五分钟的异步审阅。
The executive decision that required three levels of approval might soon happen in minutes. Companies can scale, truly scale, without the degradation we’ve accepted as inevitable.
过去需要三级审批的管理层决策，未来或许几分钟内就能完成。企业得以真正实现规模化发展，摆脱以往我们不得不接受的效率递减问题。
The second story is about steam. Factories of the early Industrial Revolution were built next to rivers, harnessed to waterwheels. When steam engines arrived, factory owners initially just replaced waterwheels with steam engines and kept everything else the same. Productivity gains were modest.
第二个故事关于蒸汽机。工业革命初期，早期纺织厂都建在河边，依靠水车驱动。当蒸汽机出现后，工厂主起初只是用水车换成蒸汽机，其他一切照旧，生产力提升十分有限。
The real breakthrough came when factory owners realized they could break free from water entirely. They built larger factories closer to workers, ports, and raw materials. And they redesigned factories around steam engines (later, when electricity became widespread, factory owners went further, abandoning central drive shafts and putting small motors on individual machines).
真正的突破，源于工厂主意识到他们可以彻底摆脱对水流的依赖。他们在靠近工人、港口和原材料的地方建造更大的工厂，并围绕蒸汽机重新设计工厂布局。后来电力普及，工厂主进一步去中心化，不再依赖中央传动轴，而是为不同机器配备独立的小型发动机。
Productivity exploded, and the Second Industrial Revolution took off. We’re still in the waterwheel phase. We’re bolting chatbots onto workflows designed for humans. We haven’t reimagined what organizations could look like when old constraints disappear—when your company can run on infinite minds that never sleep.
生产力随之爆发，第二次工业革命才真正腾飞。我们现在仍处于”替换水车”的阶段，只是把人工智能聊天机器人嫁接到现有工具上。当旧有的限制瓦解，当你的公司可以依靠永不休眠的无限心智驱动时，我们尚未重新构想组织会是什么样子。
At Notion, we’ve been experimenting. Alongside 1,000 human employees, over 700 agents handle repetitive work: taking meeting notes and answering questions to consolidate tribal knowledge, processing IT requests and logging customer feedback, helping new hires learn about benefits, writing weekly status reports so people don’t have to copy-paste. This is just the beginning. The real gains are limited only by our imagination and inertia.
在我所在的公司Notion，我们一直在进行实验。除了1000名员工外，现在有700多名智能体负责处理重复性工作。它们记录会议纪要并回答问题以整合部落知识，处理IT请求并记录客户反馈，帮助新员工了解员工福利，撰写每周状态报告，省去人们复制粘贴的麻烦。而这仅仅是起步阶段，真正的收益仅受限于我们的想象力和惯性。
Economies: From Florence to Megacities
经济体：从佛罗伦萨到超级都市
Steel and steam didn’t just change buildings and factories—they transformed cities. Until a few hundred years ago, cities were human-scaled. You could walk across Florence in 40 minutes. The rhythm of life was determined by how far a person could walk and how loud a voice could carry.
钢铁和蒸汽不仅改变了建筑和工厂，更重塑了城市。几百年前，城市都是以人为本的尺度。你可以在四十分钟内步行穿越佛罗伦萨，生活节奏由步行距离和声音传播范围决定。
Later, steel frames made skyscrapers possible. Steam engines powered railroads connecting city centers to suburbs. Elevators, subways, and highways followed. The size and density of cities exploded, giving rise to Tokyo, Chongqing, and Dallas.
后来，钢铁框架让摩天大楼成为可能，蒸汽机驱动铁路连接城市中心与周边地区，电梯、地铁、高速公路相继出现，城市的规模和密度呈爆发式增长，东京、重庆、达拉斯应运而生。
These aren’t just scaled-up Florences—they’re entirely new ways of living. Megacities are confusing, alienating, and hard to navigate. That complexity is the cost of scale. But they also offer more opportunities and freedom, with more people creating more combinations in more ways than the human-scaled cities of the Renaissance.
这些并非佛罗伦萨的放大版，而是全新的生活方式。超级都市令人迷茫、疏离，难以融入，这种复杂性是规模化的代价。但它们也提供了更多机遇和自由，更多人以更多样的方式，进行着比文艺复兴时期以人为本的城市更丰富的组合创造。
I think the knowledge economy is about to undergo the same transformation. Today, knowledge work makes up nearly half of US GDP. But most of it still operates at human scale: teams of dozens, workflows paced by meetings and emails, organizations that hit a wall once they grow beyond a few hundred people. We’ve built “Florence” with stone and wood.
我认为知识经济即将经历同样的蜕变。如今，知识工作占美国GDP近半。但其中大多数仍以人力尺度运作：数十人的团队，由会议和邮件设定节奏的工作流，一旦超过几百人就会遇到瓶颈的组织。我们用石头和木头建造了”佛罗伦萨”。
When AI agents scale, we’ll build “Tokyo.” Networks that hold thousands of agents and humans. Workflows that run nonstop across time zones, no need to wait for someone to wake up. Decision-making with just the right amount of human-in-the-loop, precisely placed.
当AI智能体规模化上线，我们将建造”东京”。容纳数千智能体与人类的协同网络。跨时区不间断运行的工作流，无需等待某人醒来。决策机制中精准嵌入了适量的人在回路。
This will feel new. Faster, higher-leverage, but also disorienting at first. The rhythm of weekly meetings, quarterly planning, annual reviews may no longer apply. New rhythms will emerge. We lose some readability. We gain scale and speed.
这将带来全新的体验。更快速、更高杠杆，但初期难免也更让人迷惑。周会、季度规划和年度评估的节奏可能不再适用。新的节奏即将诞生。我们失去了一些可读性。我们收获了规模与速度。
Beyond the Waterwheel Phase
超越水车阶段
Every generation’s miracle material forces humanity to stop driving backward and start imagining the new world. Carnegie saw steel and envisioned skylines. Lancashire mill owners saw steam and imagined factories no longer chained to rivers. We see AI and only want it to be a copilot.
每一代奇迹材料，都迫使人类停止”倒着开车”，开始想象新世界。卡内基看见钢铁，就预见城市天际线；兰开夏厂主看见蒸汽，就设想不再被河流绑住的厂房；我们看见AI，却只想让它做”副驾驶”。
Let’s stop retrofitting AI into old waterwheels. Let’s build the factories of the mind—powered by steam, reinforced with steel, and staffed by infinite minds.
让我们停止将人工智能硬塞进旧水车。让我们建造思想的工厂——由蒸汽驱动，用钢铁加固，由无限心智组成。
Steel. Steam. Infinite minds. The next skyline is there, waiting for us to build it.
钢铁，蒸汽，无限心智。下一幅城市天际线就在前方，等待我们亲手筑就。