本集简介
双语字幕
仅展示文本字幕，不包含中文音频；想边听边看，请使用 Bayt 播客 App。
所以互联网称我为网上最AI精神错乱的人之一。
So the Internet calls me one of the most AI psychotic people online.
所以，我以历史上最被取消的人之一的故事开始我的演讲，是再合适不过了。
So it's only right that I start my talk with a story about one of the most canceled men in history.
他叫巴鲁克·斯宾诺莎。
His name was Baruch Spinoza.
如果你没选修哲学课，这里是你需要知道的要点。
And in case the philosophy elective wasn't your thing, here are the highlights you need to know.
1929年，一位纽约拉比通过电报向爱因斯坦发起挑战。
In 1929, a New York rabbi challenged Einstein by telegram.
你相信上帝吗？
Do you believe in God?
请用50个字回答。
Answer in 50 words.
爱因斯坦用25个字回答了。
Einstein answered in 25.
我相信斯宾诺莎的上帝，他在世界的和谐法则中显现自己，而不是那个关心人类命运和行为的上帝。
I believe in Spinoza's God who reveals himself in the lawful harmony of the world, not in a God who concerns himself with the fate and doings of mankind.
当时最著名的科学家提出了最大的问题，指向了斯宾诺莎。
The most famous scientist alive asked the biggest question there is pointed at Spinoza.
但巴鲁克·斯宾诺莎自己的社区对他做了什么？
But here's what Baruch Spinoza's own community did to him.
阿姆斯特丹，1656年7月27日。
Amsterdam 07/27/1656.
斯宾诺莎23岁，是塞法迪犹太社区紧密团体的一员。
Spinoza is 23 years old, a member of a tight knit Sephardic Jewish community.
他站在犹太教堂里，长老们用社区有史以来最恶毒的诅咒将他逐出教门。
He stands in a synagogue while the elders excommunicate him with the most violent curse the community ever produced.
白天受诅咒，夜晚受诅咒。
Cursed be he by day and cursed be he by night.
躺下受诅咒，起来受诅咒。
Cursed be he when he lies down and cursed be he when he rises up.
没人能跟他说话。
Nobody may speak to him.
没人能跟他做生意。
Nobody may trade with him.
没人能靠近他四肘之内。
Nobody may come within four cubits of him.
没人能读他写的任何东西。
Nobody may read anything he writes.
而且，在斯宾诺莎的社区那个世纪发出的约四十条禁令中，这条禁令是唯一没有悔改条款的。
And this ban, uniquely among the roughly 40 bans Spinoza's community issued that century, has no repentance clause.
它从未被解除。
It has never been lifted.
严格来说，今天它仍然有效。
It technically is still in force today.
斯宾诺莎当时才23岁。
Spinoza was 23.
他的罪行是表达被禁止的思想的邪恶观点。
His crime was evil opinions expressing forbidden thoughts.
他的惩罚是从社区中完全除名。
His punishment was complete deletion from the community.
在他的社区诅咒他之前，他们试图每年给他一千荷兰盾。
Before his community cursed him, they tried to buy him a thousand guilders a year.
这是一笔巨款。
Serious money.
他只需要偶尔去犹太教堂露个面，然后闭嘴。
All he had to do was show up at synagogue once in a while and keep his mouth shut.
用创始人的话来说，听听这个。
Hear that in founder terms.
他们给他薪水，让他停止建造。
They've offered him a salary to stop building.
他拒绝了。
He said no.
他说，给一万也不行。
Not for 10,000, he said.
他要的是真理，不是安逸。
He wanted truth, not comfort.
就在他被逐出教会前不久，一个狂热分子拿着刀朝他冲来。
Shortly before his excommunication, a fanatic came at him with a knife.
刀刃划破了他的斗篷，但没有伤到他。
The blade tore through his cloak and missed him.
他留着那件斗篷，破口没有修补，直到去世。
He kept that cloak, scar unmended, for the rest of his life.
他想记住思想的代价。
He wanted to remember what ideas cost.
那么，这位十七世纪最被排斥的人接下来做了什么？
So what does the most canceled man of the seventeenth century do next?
他磨起了镜片。
He grinds lenses.
白天，他制作光学仪器，这些工具能让人们看得比肉眼更远。
By day, he makes optical instruments, tools that let human beings see further than their eyes allow.
他做得非常好，以至于欧洲最优秀的科学家都来找他。
He makes them so well that the best scientists in Europe seek them out.
而到了晚上，他写一本非常危险的书，以至于生前无法出版。
And by night, he writes a book so dangerous he cannot publish it while he is alive.
他44岁去世时，肺里满是制作他人镜片时吸入的玻璃粉尘，手稿锁在他的书桌里。
When he dies at 44, lungs full of glass dust from making other people's lenses, the manuscript is locked in his writing desk.
他临终的指示是，用运河驳船把书桌运到阿姆斯特丹的出版商那里。
His dying instruction, ship the desk by canal barge to his publisher in Amsterdam.
那份手稿后来成为他的遗作，立刻引起了整个欧洲的关注，并启发了启蒙时代一些最重要的哲学家。
That manuscript became his posthumous works, which attracted immediate attention across Europe and inspired some of the most important philosophers of the enlightened.
你可能会问，斯宾诺莎和创业公司有什么关系？
What does Spinoza have to do with startups, you might ask?
嗯，这个人被所有认识他的人排斥，有人出钱让他停止，还差点因为运送东西而被杀。
Well, here is a man canceled by everyone he knew, offered a salary to stop, nearly killed for shipping.
而他的回应是，白天制造精密工具，晚上独自一人写欧洲最危险的书，没有得到任何人的许可。
And his response was to build precision tools by day and write the most dangerous book in Europe by night alone with no permission from anybody.
如果你想创业，向斯宾诺莎学习会很有帮助。
If you're gonna start a startup, you could do well to learn from Spinoza.
他给支撑自己前进的动力起了个名字。
He had a name for the engine that kept him going.
Canatus。
Canatus.
意思是你的奋斗。
It means your striving.
每个生物体内持续前进、增强行动力的驱动力。
The drive in every living thing to keep going and to increase its power to act.
不是你的简历，不是你的头衔，不是你的工作，而是奋斗本身。
Not your resume, not your title, not your job, the striving itself.
这场演讲讲的是放大这种奋斗的工具。
This talk is about the tools that amplify it.
那他到底说了什么，值得让人把他干掉？
So what did he actually say that was worth deleting a man over?
他那个异端学说的核心是，上帝不是坐在宝座上的国王。
The gist of the heresy was that God is not a king on a throne.
上帝已经渗透到所有存在的事物之中。
God has spread through everything that exists.
他写道：上帝即自然。
God or nature, he wrote.
四百年后，我们在智能方面犯了类似的错误。
Four hundred years later, we are making a similar mistake about intelligence.
所有人都在等待AGI，把它当作一个单一事件，一个数据中心里的神。
Everyone is waiting for AGI as a singular event, a god in a data center.
某个公告，某个临界点，某一天天空变色。
Some announcement, some threshold, some day when the sky changes color.
所以现在我要说一个斯宾诺莎异端学说的版本，四百年后更新版。
So now I'll say a version of Spinoza's heresy updated four hundred years later.
所有人都在盯着天空，但他们等待的东西其实已经在房间里了。
Everyone is watching the sky, and the thing they're watching for is already in the room.
它看起来不像神。
It doesn't look like a god.
它看起来像基础设施，一个终端窗口，一个装满Markdown文件的文件夹，一个在你睡觉时完成的任务，散布在一切事物中，这正是斯宾诺莎告诉你要寻找的地方。
It looks like infrastructure, a terminal window, a folder of markdown files, a job that finishes while you sleep, spread through everything, which is exactly where Sminoza told you to look.
AGI不是作为一个事件到来，而是以你的智能体运行在你的上下文中、做你的工作的方式扩散开来。
AGI isn't arriving as an event, it's arriving diffused as your agent running on your context, doing your work.
我称之为个人AGI，不是为所有人同时提供的人工通用智能，而是为一个人——你——提供的通用智能。
I call it personal AGI, not artificial general intelligence for everyone all at once, general intelligence for one person, you.
这是很多人的梦想。
This was a dream of a great many people.
万尼瓦尔·布什称之为“Memex”，一种将成为你和你的大脑延伸的机器。
Vannevar Bush called it the memmex, a machine that would be an extension of yourself and your brain.
我想明确我的意思，因为“个人AI”这个词已经被营销部门用滥了。
And I wanna be precise about what I mean because the words personal AI has already been captured by marketing departments.
我不是指那种你每月付20美元的聊天机器人。
I do not mean a chatbot you pay $20 a month to.
我不是指稍微好一点的自动补全。
I do not mean a slightly better auto complete.
我不是指只知道你日历而其他什么都不懂的助手。
I do not mean an assistant that knows your calendar and nothing else.
那只是你租来的订阅服务。
That's just a subscription you rent.
那是你不拥有的企业级AGI。
It's a corporate AGI you don't own.
你关掉标签页它就重置了。
It resets when you close the tab.
它知道的都是别人已经知道的东西。
It knows what everyone else already knows.
而且当背后的公司转型时，你那所谓的助手会在别人的时间表上被做脑叶切除。
And when the company behind it pivots, your so called assistant gets a lobotomy on someone else's schedule.
个人AGI是另一种东西。
Personal AGI is a different animal.
一个在你自己的基础设施上运行的代理，读取你拥有的记忆，执行你写的程序，并且不断积累。
An agent that runs on your infrastructure, reads from the memory you own, executes procedures you wrote, and compounds.
你不拥有的企业AGI只有在公司发布新东西时才会变得更好。
The corporate AGI that you don't own gets better only when the company ships something.
你的个人AGI每天都在变得更好，因为你每天使用它，它就更了解你的生活。
Your personal AGI gets better every single day you use it because every day it knows more of your life.
其中一个是你要消费的产品。
One of these is a product you consume.
另一个是你构建的资产。
The other is an asset you build.
世界上几乎没有人拥有这第二种东西，而这个领域里的每个人到周一都可能拥有它。
Almost nobody in the world has this second thing yet, and everyone in this arena could have it by Monday.
我相信这种智能应该被拥有，被你拥有，而不是租来的。
And I believe intelligence, intelligence of this kind should be owned, owned by you, not rented.
如果你去为自己打造这个东西，2034年就不必像1984年那样。
If you go forth and build this for yourself, twenty thirty four doesn't have to be like nineteen eighty four.
你可能会问，为什么个人AGI现在才出现。
You might ask why this personal AGI is happening only now.
嗯，我认为这是因为智能体所能做的事情，而在编码智能体上这一点最为明显。
Well, I think it's because of what agents can do and nowhere is it more obvious than encoding agents.
2013年，我是YC的合伙人，晚上在搭建我们的内部社交网络Bookface。
In 2013, I was a YC partner building Bookface, our internal social network at night.
我每天大概只写出14行有用的代码，如果你了解程序员生产力的文献，这正好是中位数。
I shipped maybe 14 useful lines of code a day, which if you know the literature on programmer productivity is dead on median.
那是我全力以赴的状态。
That was me at full effort.
今年，我全职运营YC，同样的脑子，同样的时间，外加早上5点接孩子。
This year, I run YC full time, same brain, same hours, plus a 05:00 kid pickup.
我算了一下我的产出，大概是2013年的400倍。
I did the math on my output, and I'm at about 400 x what I did in 2013.
在第三排的怀疑者给我泼冷水之前，我先自己泼一盆。
Now before the skeptic in row three deflates that number for me, let me deflate that for myself.
你不相信那些原始代码行数？
You don't trust the raw lines of code?
好吧。
Fine.
就算你用最苛刻的冗余惩罚来算，假设代理写的代码很臃肿。
Apply the most pathological verbosity penalty you can stomach and assume the agent writes bloated code.
假设一半都是脚手架代码。
Assume half of it is scaffolding.
假设我在自我吹嘘，这总是有可能的。
Assume I'm flattering myself, which is always a live possibility.
但即使是最低估计也有八倍，中位数估计则是十倍以上。
It's still eight x at the absolute floor and 10 times that in the middle of the range.
不管你怎么算，这个数字都很大。
The number is large no matter how you torture it.
这还只是代码方面。
Now this is just code.
如果你刚入行，那你运气不错。
And if you're at the beginning of your career, you're in luck.
这同样适用于设计。
This applies to design.
也适用于产品管理。
This applies to product management.
也适用于增长。
This applies to growth.
它适用于你可能想做的每一件事。
This applies to every part of what you might want to do.
编码的乘数效应不仅仅适用于编码。
The multiplier for coding is not just for coding.
它适用于所有知识型工作。
It's for every piece of knowledge work.
而且不只是我这么认为。
And it's not just me.
在YC，我们能以投资组合的规模看到这一点。
At YC, we get to watch this at portfolio scale.
一年半前，也就是25年冬季那批，有四分之一的公司的代码库有95%是AI生成的。
A year and a half ago in the winter twenty five batch, a quarter of the companies had code bases that were 95% AI generated.
那些公司现在什么都用AI代理，不只是代码。
Those companies use AI agents for everything now, not just code.
而那批公司有望成为YC历史上增长最快、最赚钱的批次之一。
And that batch is on track to becoming one of the fastest growing, most profitable batches in the history of YC.
我知道相关性和因果性的区别，所以我要谨慎地说。
Now I know what a correlation is, so let me say it carefully.
我无法证明AI生成的代码和其他一切导致了增长。
I cannot prove that the AI generated code and everything else caused the growth.
但我能告诉你的是，我们资助的增长最快的创始人并没有把AI当作自动补全工具。
But what I can tell you is that the fastest growing founders we fund are not treating AI as autocomplete.
他们把它当作一支劳动力队伍。
They are treating it as a workforce.
有两倍的人和一百倍的人在使用同一个Claude，同样的权重，同样的上下文窗口大小，同样的API。
There are two x people and there are 100 x people who are using the same Claude, same weights, same context window size, same API.
但杠杆不在于权重。
But the leverage is not in the weights.
而在于你给它什么上下文，它有多相关，以及它是否在正确的步骤发生。
It's in what context you give it, how relevant it is, and does it happen at the right step.
我们稍后再回到这个话题。
We'll come back to this.
现在，斯宾诺莎有一个定义，我每周都会想到。
Now, Spinoza has a definition I think about every single week.
在《伦理学》中，他将喜悦定义为你的行动能力增强的感觉，这就是为什么当一个代理第一次在一个下午完成你一周的工作时，它感觉不像是一种便利。
In ethics, he defines joy as the feeling of your power of acting increasing, which is why the first time an agent does a week of your work in an afternoon, it doesn't feel like a convenience.
它感觉像是喜悦。
It feels like joy.
这可不是我在打比方。
And that's not me being poetic.
这是专业术语。
That's the technical term.
你的行动力增强了。
Your power of acting increased.
你的生命力变强了。
Your canatus just got bigger.
顺便说一下，他给相反的感觉下了定义，悲伤，就是你的行动力减弱的感觉。
He defined the opposite to you, by the way, sadness, the feeling of your power of acting decreasing.
如果你周日晚上的心情特别沉重，感觉你对世界的影响力在减弱，觉得自己在消极怠工，那你感受到的就是这个。记住这种感觉，因为我们会在演讲的后半部分再回到这个话题。
If your Sunday nights have a specific heaviness, like your ability to influence the world is receding, that you feel like you're quiet quitting, then this is what you feel and hold that thought because we're coming back to this in the second half of this talk.
而且这还会涉及政治。
And it gets political.
所以，这就是你未来十年人生的方程式。
So here's the equation for the next decade of your life.
一个前沿模型，它是租来的、像商品一样，而且每个季度都在降价，再加上你的上下文，那是你独有的。
A frontier model, which is rented and a commodity and getting cheaper by the quarter, plus your context which is owned by you and unique.
不。
No.
而且理想情况下，地球上没有其他人拥有它。
And ideally nobody else on this earth has it.
再加上一个把它们连接起来的框架。
Plus a harness that wires them together.
那个框架可能是Open Claw、Hermes agent、Claude code或者Codex。
That harness might be Open Claw, Hermes agent, Claude code, or Codex.
把这些加起来，你就得到了一个像你一样行动但速度非常快的代理。
Add that up, and that gives you an agent that acts like a very fast version of you.
模型质量是租来的，但你的大脑理想情况下是你自己拥有的。
Model quality is rented, but your brain is owned ideally by you.
马歇尔·麦克卢汉说过，技术是人的延伸。
Marshall McLuhan said that technology is an extension of man.
史蒂夫·乔布斯曾把电脑称为大脑的自行车。
Steve Jobs called a computer a bicycle for the mind.
如果你拥有我描述的东西，那你就拥有了一辆自动驾驶的火箭。
And if you have what I'm describing here, then you have a self driving rocket.
保罗·格雷厄姆教会了这栋楼里每一位创始人两件事。
Paul Graham taught every founder in this building two things.
做出人们想要的东西，以及做那些无法规模化的事情。
Make something people want and do things that don't scale.
这两条至今仍然主导一切。
Both still govern everything.
新的变化在于第二条的倍增效应。
What's new is the multiplier on the second one.
现在，一位创始人可以通过智能体，以规模化的方式去做那些无法规模化的事情。
Agents are how one founder now does unscalable things at scale.
建议没有变，但所有初创公司的物理法则，以及你能做到的事情，都变了。
The advice didn't change, but the physics of all startups and of what you can do did.
斯宾诺莎磨镜片，那些透镜让人们能看到眼睛极限之外的东西。
Spinoza ground lenses, instruments that let people see past the limits of their eyes.
我想用接下来的十五分钟，向大家展示为思维磨镜片是什么样子。
I wanna wanna spend the next fifteen minutes showing you what grinding lenses for the mind looks like.
这是我实际用来管理生活的机制，而且每个概念都能应用到你所使用的任何技术栈中。
This is the machinery I actually run my life on, and every concept travels to whatever stack you use.
我们先从工作记忆说起，因为它解释了一切。
Let's start with working memory because it explains everything.
你我作为人类，脑子里一次能记住大约七件事。
You and I, as human beings, hold about seven things in our head at once.
七加减二。
Seven plus or minus two.
这是认知心理学中最著名的一篇论文。
It's the most famous paper in cognitive psychology.
这就是为什么本地电话号码是七位数，也是为什么你会忘记购物清单上的第八项。
It's why local phone numbers are seven digits and why you forget the eighth item on a grocery list.
这就是人类全部的短期记忆容量。
That is the entire working memory of a human being.
人类建立的每一个机构，每一张清单，每一张组织架构图，每一个文件柜，每一次站会，都是为了弥补这个限制而造的。
And every institution humanity has ever built, every checklist, every org chart, every filing cabinet, every stand up meeting is a prosthetic for that limit.
然而，一个AI智能体却能处理一百万个token。
An AI agent though holds a million tokens.
那大概相当于一千页纸。
That's about a thousand pages.
相当于三本《哈利·波特》同时摊开在它脑子里。
Three Harry Potter books sitting open on its head all at once.
而且它能在任何一本里找到一根针，并在几秒内综合三本书的内容。
And it can find a needle in any of them and synthesize across all three in seconds.
三本《哈利·波特》对比七个数字。
Three Harry Potter books versus seven digits.
你可以说这还不完全是AGI，但这已经是一种不同的运行模式了。
You could argue that's not quite AGI yet, but it is already a different operating regime.
而且地球上几乎每个人仍然在按照为七位数大脑设计的组织架构和做事方式来生活。
And almost everyone on Earth is still running their life on an org chart and a way of doing things designed for the seven digit brain.
把这个数字往另一个方向想想。
Run that number in the other direction.
一千页很多，但也非常少。
A thousand pages is a lot, but it is also very little.
你的人生不是三本书。
Your life is not three books.
你的人生是一座图书馆。
Your life is a library.
你发过的每一封邮件，每一次会议，每一个决定以及背后的原因，和每个认识的人的每一次对话。
Every email you ever sent, every meeting, every decision, and every reason behind it, every conversation with every person you know.
决定你的智能体是天才还是金鱼的问题是这样的。
The question that determines whether your agent is a genius or a goldfish is this.
谁来决定，或者什么来决定，桌上打开的是哪三本书。
Who decides or what decides which three books are open on the desk.
这就是大脑的本质。
And that's what a brain is.
这就是g brain的初衷——既是图书馆，又是图书管理员。
That's what g brain is meant to be, the library plus the librarian.
我一直在公开构建g brain。
I've been building g brain in the open.
我的个人Open Claw有一个Karpathy风格的知识维基，大约有22万个Markdown页面。
My personal Open Claw has a Karpathy style knowledge wiki with about 220,000 markdown pages.
记录了我25年的生活。
Twenty five years of my life diarized.
每一封邮件、每一次会议、我的笔记、我的照片、我的草稿，还有我犯过的错误。
Every email, every meeting, my notes, my photos, my drafts, the things I got wrong.
大部分由智能体整理，由智能体策展，由智能体搜索。
Compiled mostly by agents, curated by agents, searched for by agents.
但我绝不会重复问已经回答过的问题。
But I never re ask a question I already answered.
而系统里的真实体验才是关键。
And the lived experience in the system is the point.
一位创始人给我发邮件说遇到危机，我还没读完邮件呢。
A founder emails me about a crisis before I finish reading the email.
我的代理就已经把所有我和那位创始人之前的对话都拉出来了。
My agent has already pulled every prior conversation I've had with that founder.
还有三家投资组合公司遇到过同样的问题，以及对他们真正有效的解决办法。
Three portfolio companies that hit the same wall and what actually worked for them.
我的代理做任何事时，都会利用我所知道的一切。
When my agent does anything, it does knowing everything I know.
这就是助手和同事之间的区别。
And that's the difference between an assistant and a colleague.
让我带你过一遍真实的一天，因为这比架构图更重要。
Let me walk you through an actual day because that matters more than an architecture diagram.
昨晚我睡觉的时候，我的代理处理了我的收件箱。
While I slept last night, my agent processed my inbox.
不是整理，是处理。
Not sorted it, processed it.
它知道哪些邮件来自遇到麻烦的创始人，哪些来自想向我推销的人，还有哪些来自我从未真正退订的17个邮件列表。
It knows which emails are from founders in trouble, which are from people trying to sell me something, and which are from the 17 mailing lists I never quite unsubscribed from.
重要的邮件会被分类，并附上背景信息，比如这个人是谁、我和他们的全部历史、他们字里行间真正想要什么，以及这可能对我意味着什么。
The ones that matter are triaged with context pulled from the library, who this person is, my whole history with them, what they're really asking under what they wrote, and what that might mean for me.
我醒来看到的是简报，而不是一堆邮件。
I wake up to a briefing, not a pile of emails.
每次会议前，都会有一份准备文档，包括我要见谁、我们上次说了什么、之后有什么变化，以及我应该问什么。
Before every meeting, a prep doc, who I'm meeting, what we said last time, what changed since, and what I should ask.
我午夜好奇的研究，到早上就完成了。
Research I was curious about at midnight is finished by morning.
当世界上发生有趣的事情时，我的代理通常已经读过，并对照我关心的事情进行交叉引用，在我喝咖啡之前就归档好了。
And when something interesting happens in the world, my agent has usually read it cross referenced against what I care about and filed it before I've had coffee.
在这个知识库之上，是我的代理编码框架g stack，现在有12.3万颗星，这使它进入了GitHub历史上最受欢迎的开源项目前100名。
On top of this library sits my agent coding framework g stack, a 123,000 stars now, which put it in the top 100 open source projects in the history of GitHub.
那这个完整架构的亮点到底是什么呢？
And what's actually in the punchline of this full architecture?
主要是技能文件，外加一个智能体可以操控的浏览器。
It's mostly skill files plus a browser that the agents can drive.
几页英文说明，再加上一个能对世界采取行动的方式。
Pages of English and a way to act on the world.
是Markdown，不是魔法。
Markdown, not magic.
先有丰富的技能，再有框架。
Fat skills, then harness.
让我给你看看技能文件长什么样，因为我一直提这个词，我想让你明白它其实没什么神奇的。
Let me show you what a skill file is because I keep saying this phrase, and I want you to see how unmagical it is.
这里有一个真实的例子，稍微做了点脱敏处理。
Here's a real one, lightly redact is.
上面写着：当Circle Back的会议录音到达时，用说话人标签转写，提取出做出的承诺、承诺人以及截止日期。
It says, when a meeting recording lands from circle back, transcribe it with speaker labels, pull out the commitment made, who made it, and the deadline.
把提到的每个人跟资料库核对一下，然后链接到他们的页面。
Cross check every person named against the library and link their pages.
把摘要归档到这里，完整记录放到那边。
File the summary here, full transcript there.
如果有什么跟我们已经相信的内容矛盾，就标记出来。
If anything contradicts something we already believe, flag it.
不要覆盖它。
Don't overwrite it.
就这样。
That's it.
这就是一个技能。
That's a skill.
就是一页英文。
It's a page of English.
一个聪明的实习生，其实只要会读的人都能照着做。
A smart intern, anyone really who could read could follow it.
这其实就是个测试。
And that's the test actually.
如果一个聪明的实习生能照着做，那智能体也能执行，这其实挺深刻的。
If a smart intern could follow it, an agent can run it, which means actually a kind of profound thing.
我知道我这么说招了不少批评，但我现在比以往任何时候都更确信这一点，尤其是现在。
I know I caught a lot of flack for talking about this, but I think it's more true than ever, especially now.
Markdown 其实就是代码。
Markdown is actually code.
如果你能用英语写出清晰的指令，那你就是程序员。
If you can write clear instructions in English, you're a programmer.
编译器就是语言模型。
The compiler is a language model.
所以这不再只是工程师的事了。
And that's why it's not just for engineers anymore.
在 YC，我们的媒体人员、活动人员、财务团队，那些一辈子都没打开过终端的人，都在创建技能文件和定时任务。
At YC, our media people, events staff, finance team, people who never open a terminal in their lives are building skill files and scheduled jobs.
我们财务部有个人，用内部代理工具把大约100个Excel工作簿整合成了一个应用。
One of our finance folks compiled about a 100 Excel workbooks into a single app she built with an internal agent.
她不是程序员。
She is not a programmer.
现在她是代理经理了。
She is a manager of agents now.
每个人很快都会这样。
Everyone is about to be.
这里最关键的问题是：计算发生在哪里？
The most important question to ask here is where is the computation happening?
答案只有两个，而混淆它们会导致我见过的所有代理失败。
And there are exactly two answers, and confusing them causes every agent failure I've ever seen.
有些计算属于潜在空间。
Some computation belongs in latent space.
品味、判断力、从模糊请求中理解人类真正想要什么，这些都在模型里，你用markdown文件来引导它。
Taste, judgment, reading what a human actually wants from a vague request, that lives in the model and you steer it with a markdown file.
然后有些计算属于确定性空间，比如算术、SQL查询。
And then some computation belongs in deterministic space, the arithmetic, the SQL query.
比如说，你今天要参加哪些分会场的座位安排。
For instance, the seating chart for what sessions you're going to go to today for your breakouts.
所有这些都需要存储在SQL数据库中，供markdown文件使用。
All of that needs to be stored in a SQL database used by the markdown files.
在这方面聪明一点会大有帮助。
Being smart about this goes a long way.
让一个代理或人安排五个人围坐在一张桌子旁。
Ask an agent or human to seat five people around a table.
这很容易。
That's easy.
在潜在空间里做。
Do it in latent space.
让它像我们刚才为你做的那样，为竞技场里的6000人制定个性化日程，你的潜在空间代理就需要编写一些代码来跟踪这些。
Ask it to make custom schedules for 6,000 people in an arena like we just did for you, and your latent space agent needs to write some code to keep track of it.
你参加这次会议的经历，就得靠 Markdown 文件以这种方式调用代码。
Your experience at this conference had to be markdown files calling code in exactly this way.
没有代码，你根本做不到。
And you can't do it without the code.
模型失败的地方，也正是我们失败的地方。
The model fails where we fail.
解决办法就是让模型像人类一样计算。
The fix is having the model compute the way humans compute.
潜在空间和确定性空间，Markdown 文件调用数据库和脚本。
The latent and the deterministic, Markdown files calling databases and scripts.
简单，但一切其实都建立在这上面。
Simple, but it's what everything is actually built on.
我再给你一个证据，这是我最喜欢的一个，因为你现在就身处其中。
And I'll give you one more receipt, my favorite one because you're sitting inside it right now.
五天前，我决定这次演讲要提到斯宾诺莎，我最喜欢的哲学家之一，尤其是因为他当时被抵制得那么厉害。
Five days ago, I decided this talk needed Spinoza, one of my favorite philosophers, especially because of how canceled he got.
所以我的智能体去搞来了三本关于这个人的最好的传记，作者是纳德勒、戈尔茨坦和斯图尔特，总共大约1500页。
So my agent went and acquired three of the best biographies about the man, books by Nadler, Goldstein, and Stuart, about 1,500 pages.
它把这三本全读完了。
It read all three.
它给我做了一份综合整理，包括他生平的年表，三位传记作者之间所有意见分歧的地方，还有最好的原话引用，都附上了章节出处。
It built me a synthesis, a dated chronology of his life, every place the three biographers disagree with each other, and the best verbatim quotes with chapter citations.
而且因为它知道我需要什么，它还把他一生中最适合讲述的十个时刻按优先级排好了，还附上了讲述建议，比如那次持刀袭击、那次贿赂、那张书桌。
And because it knows what I need, the 10 most tellable moments of his life ranked with delivery notes, The knife attack, the bribe, the desk.
我们开场时那些可能让你在二十分钟前感到震撼的每一个瞬间，都来自那次通宵的运行。
Every beat of our opening that might have given you some chills twenty minutes ago came out of that overnight run.
1500页变成了一篇可以直接上台讲述的故事，而且我还能自己修改。
1,500 pages became a stage ready story that I could edit.
我管这叫“汇编技能”，我每天都在用。
I call it a compendium skill, and I use it daily.
这是一种个人技能，是深度研究的超级加强版，比任何企业AI产品给你的都要深入得多。
It's a personal skill that is a mega mega version of deep research, only deeper than anything the corporate AI products will give you.
你现在看的这场演讲的主线，其实就源于我们正在说的这个机器。
The spine of this talk you're watching was inspired by the machine we're describing now.
如果你想知道我的知识库最初是怎么来的，那可不是从22万页开始的。
And if you're wondering where my mine actually started, it was not 220,000 pages.
最开始只是一个文件夹。
It was a folder.
里面有几个Markdown文件，记录着我合作的公司和经常联系的人。
It was a few markdown files about the companies I was working with and the people I kept emailing.
这个知识库变大，和其他任何东西变大的方式一样。
And the library got big the same way anything gets big.
每天积累一点点，再加上智能代理帮忙整理归档。
A little everyday compounding with agents doing the filing.
没人会先建仓库。
Nobody builds the warehouse first.
你得先搭起一个架子。
First, you build one shelf.
今晚你坐下来和智能体一起工作时，你并不是在编程。
When you sit down with an agent tonight, you're not coding.
你是在管理一支由Markdown构成的劳动力队伍。
You're managing a workforce made of markdown.
一个技能文件就是一个员工。
A skill file is an employee.
它有一项能力，一项工作，写得清清楚楚，让新来的人也能执行。
It has one capability, one job written down clearly enough that someone new could execute it.
一个解析器就是一张组织结构图。
A resolver is an org chart.
任务进来后，由它决定哪个Markdown文件或谁来处理。
A task comes in and it decides which markdown file or who handles it.
这意味着，在你注册公司之前，在你找到联合创始人、设计Logo或制作演示文稿之前，你就能运营一个组织了。
Which means that before you ever incorporate anything, before you have a co founder or a logo or a deck, you can already be running an organization.
一个由你加上你的智能体组成的组织。
An organization of one plus your agents.
你就是创始人，也是“你公司”的整个管理层，你手下的员工数量现在由你决定。
You are the founder and the entire management layer of You Incorporated and the headcount under you is now whatever you decide it is.
这已经催生出打破旧有模式的公司。
This already produces companies that break the old math.
我们2024年夏季班中涌现出的公司，从公开发布到九位数营收只用了八个月。
Emergent out of our summer twenty four batch went from public launch to 9 figures of revenue in eight months.
当年化收入突破1500万美元时，他们只有15个人。
When they crossed $15,000,000 in annualized revenue, they were 15 people.
零售领域的2024年冬季班，以大约40人实现了6000万美元的年化收入。
Retail, winter twenty four, hit 60,000,000 annualized with about 40.
这样的人均营收以前从未有过，无论是软件、石油还是铁路行业。
That revenue per person did not exist before, not in software, not in oil, not in railroads.
这些并不是异类。
And these aren't freaks of nature.
他们是首批基于新规则原生构建的公司，而每一家都始于一两个人，按照我刚才描述的方式运作。
They're the first companies built natively on the new physics, and every one of them started as one or two people wired the way I just described.
现在想象一下我们在狗镇的那个批次房间，每天都有数百位创始人，每个人都在做以前一个人一整年的工作量。
Now picture our batch room in the dog patch, Hundreds of founders every single day, each one of them doing what used to be a person's entire year of work.
这不是未来。
That is not the future.
这就是现在这个批次的基准。
That is the bar right now with this batch.
如果你不这么做，你的竞争对手会，他们会礼貌地抢走你的生意，还谢谢你。
If you're not doing it, your competitor is, and they will eat your lunch politely and thank you for it.
这也改变了软件的本质。
It also changes what software even is.
软件不再需要那么珍贵了。
Software doesn't have to be precious anymore.
你可以在一个周末内为一个人群构建出你需要的工具。
You can build exactly the tool you need for the audience of one in a weekend.
过去的建议是挠自己的痒处，希望它能成为一个市场。
The old advice was scratch your own itch and hope it's a market.
新版本好多了。
The new version is much better.
挠自己的痒处，因为挠痒几乎不花钱，而且你为自己做的工具，有些可能会变成整个公司。
Scratch your own itch because scratching itches is nearly free, and some of your tools for one will turn out to be entire companies.
你会知道的，因为别人会开始求着要这些工具。
You'll know because other people start begging for them.
在讲怎么做之前，我得先坦白一点，免得你抓我什么把柄。
And one honest caveat before the how to because you catch me out in any way.
没人打理的大脑，就是个搜索功能强大的垃圾场。
A brain nobody curates is a garbage dump with great search.
检索会自信地翻出一个过时的事实。
Retrieval will surface a stale fact with total confidence.
一个糟糕的技能文件会把糟糕的流程永远固化下来。
A bad skill file encodes a bad process forever.
所以，基础就是记忆加上卫生习惯：每个事实都要有来源，新信息和旧信息冲突时要检查矛盾，还要有一个真正负责修剪的图书管理员。
So the primitive is memory plus hygiene, provenance on every fact, contradiction checks when new information collides with old, and a librarian whose actual job is pruning.
把大脑当作生产基础设施来对待，它就会产生复利效应。
Treat the brain like production infrastructure and it compounds.
如果把它当作垃圾场，你就会得到一个非常自信但错误百出、且无法追溯的智能体。
Treat it like a dumping ground, and you get a very confident agent that is wrong in ways nobody can trace.
到目前为止，讲的都是理念和实例。
Everything so far is philosophy and receipts.
那我们就进入正题，讲讲具体怎么做。
So let's get into some how to.
如果你按照我接下来六分钟讲的内容去做，你会领先于99%看了这场演讲却只是点头的人。
If you do what I describe in the next six minutes, you'll be ahead of 99% of people who watched this talk and just nodded.
今晚第一步，选一个工具框架，在你自己的机器上运行一个智能体。
Step one tonight, pick a harness and run an agent on your own machine.
我用的是OpenClaw和Hermes智能体，搭配g brain。
I use OpenClaw and Hermes agent with g brain.
这个的托管版本在gbrain.io上。
A hosted version of this is at gbrain.io.
它是免费的。
It's free.
G brain 本身是免费且开源的。
G brain itself is free and open source.
我总是推荐法拉利。
I always recommend the Ferrari.
但说实话，本田也很不错。
But I'll be honest, the Honda is really good too.
Codecs、Claude code，随便哪个都行。
Codecs, Claude code, whatever.
任何一个都能完成 99% 的工作。
Any of them will do 99% of this.
不用法拉利的好处是，你也能到达目的地，而且不用经常停下来在路边修车。
And the upside of not Ferrari is that it will also get you to your destination with a little less of less getting out to fix it on the side of the road.
关键是概念，而不是某个特定的仓库或产品。
The concepts are the point, not any given repo or product.
智能潜力巨大，而且有很多途径可以挖掘。
The intelligence is untapped, and there are many paths.
这个周末的第二步，开始建立你的知识库。
Step two this weekend, start your library.
不是要建一个庞大的档案库。
Not a grand archive.
就是一个存放 Markdown 文件的文件夹，导出你的笔记，如果可以的话，也导出你的邮件。
One folder of markdown files, export your notes, export your email if you can.
为你正在做的每个项目和合作的每个人各写一页。
Write one page about each project you're working on and each person you work with.
在这些页面上，写下你真正了解的事情：你们在共同构建什么，他们关心什么，你欠他们什么，他们上次说了什么。
And on those pages, write the things you actually know, what you're building together, what they care about, what you owe them, what they said last time.
这些内容，世界上没有任何模型能拥有，因为它们只存在于你的脑海中。
That stuff no model on the earth no model on earth has because it only exists in your head.
而你的大脑，正如我们之前所说，只能记住七件事。
And your head, as we established, only holds seven things.
展开剩余字幕（还有 227 条）
当智能体第一次用你的资料而不是互联网上的信息来回答问题时，你会感到那种豁然开朗，然后就再也回不去了。
The first time an agent answers a question using your context instead of the Internet's, you'll feel the click and you won't go back.
你们每个人都坐拥五到十年的个人历史，散落在各个邮箱里。
You're all sitting on you are all sitting on five, ten years of your own history in one inbox or another.
那就是你的护城河，就那样躺在那里，没有被索引，毫无用处。
That's your moat just lying there, unindexed, doing nothing.
你和这套完整架构之间，大概只隔着二十四小时。
The only gate between you and this entire architecture is probably twenty four hours.
第三步，写你的第一个技能文件。
Step three, write your first skill file.
选哪个任务很容易。
Picking it is easy.
你知道，你每周最讨厌做的那件事是什么？
You know, what's the task you do every single week that you hate the most?
可能是报销单、会议记录、每周状态更新，或者竞品调研。
Might be expense reports, meeting notes, the weekly status update, competitor research.
把它解释给你的智能体听。
Explain it to your agent.
你想做什么？
What do you wanna do?
用大白话，就像你向一个聪明的朋友解释他第一天上班要做的事那样。
In plain English, the way you'd explain to a smart friend on their first day of a job.
然后让它犯错。
And then let it get it wrong.
如果它做错了，就纠正它。
If it gets it wrong, correct it.
每一条规则、每一个例外、还有那些“哦，对了”的补充，都写进去，它就会修正。
Every rule, every exception, every oh, and also, put it in there, and it'll fix it.
那个页面现在就是一个员工了。
That page is now an employee.
运行它。
Run it.
第四步，把它设置成定期任务。
Step four, wire it up to be a recurring job.
也许就是你第三步创建的那个任务。
Maybe it's the job you just created in step three.
每天早上七点，做这个。
Every morning at seven, do this.
每周五，总结那个。
Every Friday, summarize that.
第一次醒来发现工作在你睡觉时完成了，你的思维会发生永久性的转变。
The first time you wake up to work that finished while you slept, something shifts in your head permanently.
从那天起，一天不再是你的工作单位。
That's the day that the day stops being the unit of work for you.
它变成了你能想象到的东西，应该由你的目标和你想在这个世界上创造的东西来驱动。
It becomes what you can imagine, and it should be driven by what your goals are and what you wanna create in the world.
第五步，这是区分复利者和浅尝辄止者的纪律。
Step five, this is the discipline that separates the compounders from the dabblers.
永远不要做一次性工作。
Never do one off work.
大多数人运行一个操作，用一个代理，然后就把上下文扔掉了。
Most people run one operation with one agent and then throw the context away.
他们关掉窗口。
They close the window.
就这样。
That's it.
别这样。
Don't.
每项任务结束时，让代理把它做的事情技能化。
At the end of every task, ask the agent to Skillify what it did.
技能化是你在g brain里能找到的一个特殊技能。
Skillify is a special skill you can find in g brain.
你可以指向那个仓库，说，提取技能化，学习怎么做。
You can point it at that repo and say, extract Skillify, learn how to do it.
把它变成一个你可以永远反复使用的Markdown文件。
Turn it into a markdown file you can use and reuse forever.
我会用我在YC的说法来说。
I'll say it the way I say it at YC.
如果你需要问两次才能得到答案，那你就失败了。
If you have to ask for something twice, you failed.
那些把学到的东西记录下来的人，每天都在变得更聪明。
The person who captures what they learn gets smarter every single day.
而那些每天早上醒来就像失忆一样的人，那真是浪费时间。
The person who wakes up every morning with amnesia, well, that that's a waste of your time.
而且，如果你不能把模型变成真正的记忆，那模型再好也没用。
And it sort of doesn't matter how good the model gets if you can't turn it into real memory.
做到这五件事，我就能告诉你接下来九十天会是什么样子。
Do those five things and I can tell you what your next ninety days look like.
第一周，说实话，它就是个玩具。
Week one, honestly, it's a toy.
资料库还很薄。
The library's thin.
技能还比较生疏。
The skills are clumsy.
你修复的时间比节省的时间还多。
You're fixing more than you're saving.
第四周，飞轮开始转动。
Week four, the flywheel catches.
智能体开始用你的上下文来回答。
The agent starts answering with your context.
早上的任务产出了你真正会读的东西，而且你写了第三个和第四个技能，因为前两个奏效了。
The morning job produces something you actually read, and you write your third and fourth skill because the first two worked.
第十二周，你有了一个资料库，你还没问完它就能回答。
Week 12, you have a library that answers before you finish asking.
十几个技能文件处理着你过去害怕的那部分每周工作，还有一两个工具别人总想借用，在这个房间里，这就叫创业公司。
A dozen skill files running the parts of your week you used to dread, and one or two tools that other people keep asking to borrow, which in this room is called a startup.
这条曲线和任何复利增长的东西一样。
The curve is the same curve as any compounding thing you've ever seen.
平平平，然后突然起飞。
Flat, flat, flat, then not.
大多数尝试的人会在第二周放弃，这正是为什么坚持到第十二周的人感觉像开了挂。
Most people who try this will quit this in week two, which is precisely why the ones who don't feel like they're cheating by week twelve.
现在我得告诉你们不那么有趣的部分，因为我教的一切都是双刃剑。
Now I need to tell you the part that isn't fun because everything I taught you just cuts both ways.
我之前提到斯宾诺莎对悲伤的定义，就是感到自己行动力在减弱。
I told you Spinoza's definition of sadness earlier, the feeling of your power acting power of acting decreasing.
我说过这会变得政治化。
And I said it gets political.
就是这里。
This is where.
技能文件不是文档。
A skill file is not a document.
技能文件不是一份文档，而是你认知的一部分，是你做事的方法，从你脑子里提取出来，写下来，并且可以执行。
It's a piece of your cognition, how you do the thing, extracted from your head, written down, and executable.
你教给智能体的每一个技能，都是你自身的外化。
Every skill you teach an agent is you, externalized.
而同一个文件，根据谁控制它，会导向两种截然不同的未来。
And the exact same file is two opposite futures depending on one variable, who controls it.
举个虚构的例子，比如一位技术支持工程师。
Take a fictional example of a support engineer.
我们叫她玛雅。
Let's call her Maya.
在两年多的时间里，玛雅教会了她的智能体40项技能。
Over two years, Maya teaches her agents 40 skills.
比如凌晨两点如何对P0级故障进行分类处理。
How to triage a p zero at two in the morning.
比如如何安抚那些即将流失的客户。
How to deescalate the customer who's about to churn.
怎么写一份真正能防止下次事故的事后报告。
How to write a postmortem that actually prevents the next incident.
40个文件。
40 files.
这就是她的判断力。
That's her judgment.
她花两年时间积累的东西，就放在磁盘上。
The thing that took her two years to build sitting on a disk.
第一版，这些文件放在Maya的仓库里。
Version one, those files live in Maya's repo.
她换了工作。
She changes jobs.
这些文件跟着她走。
They go with her.
到新公司的第一天，她就能随时调用多年积累的判断力。
Day one at a new company, she's operating with years of compounded judgment on tap.
她每工作一年，经验就积累一年。
Every year she works, she compounds.
这就是主人翁精神。
That's ownership.
如果她想开一家做这个的公司，这是她的专长，而且事实证明她确实可以。
And if she wanted to start a company that does this, it's her expertise, and it turns out she can.
如今，整个初创公司可能就靠几个markdown文件。
Entire startups these days will be markdown files.
第二种情况，这些文件放在公司的代码库里，受公司IT政策管理。
Version two, those files live in the company's repo under the company's IT policy.
玛雅离开时什么也带不走。
Maya leaves with nothing.
公司继续用她的判断力运转，即使她不在。
The company keeps running her judgment without her.
40个文件永远执行，而她的名字甚至都不在提交历史里。
40 files executing forever, and her name isn't even in the commit history.
她并没有事业。
She didn't have a career.
她只是被剥削了。
She had an extraction.
同样的文件，同样的玛雅，只是一个变量不同。
Same files, same Maya, one variable.
所以这就是原则，我希望你明天能复述出来。
So this is the doctrine, and I want you to be able to repeat it tomorrow.
我相信技能文件是属于你自己的。
I believe skill files are yours.
要掌握自己的技能，因为如果你不这样做，你的工作就会变成一份技能文件。
Own your skills, because if you don't, your job becomes a skill file.
这种事情以前也发生过。
And this happened before.
工匠们拥有自己的工具。
Craftsmen owned their tools.
这就是让他们自由的原因。
That's what made them free.
工厂打破了这一点。
The factory broke that.
织布机属于工厂。
The loom belonged to the mill.
知识工作者以为我们很安全，因为我们的工具在脑子里，没人能没收。
The knowledge workers assumed we were safe because our tools lived in our heads where nobody could confiscate them.
技能文件终结了这种想法。
Skill files end that.
历史上第一次，你的认知可以被提取、存储、版本化，并被占有。
For the first time in history, your cognition can be extracted, stored, versioned, and owned.
唯一的问题是，被谁占有？
The only question is, by whom?
还记得那一千荷兰盾吗？
Remember do you remember the thousand guilders?
那个提议从来没有消失过。
That offer never went away.
它只是换了层皮。
It got rebranded.
每一个让你觉得舒服的安排，只要你的判断力在别人的地盘上增值，本质上就是每年一千荷兰盾，让你按时出现、保持沉默、别再搞自己的东西。
Every comfortable arrangement where your judgment compounds in someone else's repo is a thousand guilders a year to show up, keep quiet, and stop building your own thing.
这就是为什么你应该去创业，因为这样你才能真正让那些技能文件为你所用。
And that's why you should start a startup because this is how you can actually make those skill files work for you.
斯宾诺莎也面对过升级版的这种情况。
Spinoza faced the upgraded version too.
1673年，海德堡大学给这位被诅咒的异端提供了一个正教授职位，有薪水、有合法性、有教席，还号称有“哲学思考的自由”，前提是他不能扰乱既定的宗教。
In 1673, Heidelberg offered the cursed heretic a full professorship, salary, legitimacy, a chair, and quote freedom of philosophizing provided he not disturb the established religion.
他的回答是：我不知道那种哲学思考的自由到底有什么限制。
His answer was, I do not know what the limits of that freedom of philosophizing might have to be.
他读完了服务条款，然后拒绝了这次收购。
He read the terms of service and he declined the acquisition.
他有一个说法来形容他所保护的东西，那就是靠自己的力量，而不是靠别人的力量。
He had a phrase for what he was protecting, under your own power, as opposed to under someone else's.
无论哪种方式，你的行动力都存在。
Your power of acting exists either way.
1673年和2026年的政治问题是：谁掌控它？
The political question in 1673 and in 2026 is who commands it?
个人AI就是掌控自己的认知能力，保护自己。
Personal AI is about controlling your own cognitive abilities and protecting yourself.
这就是整个演讲的核心，一句话概括。
That's the whole thesis of this talk in one sentence.
个人AGI是让你在智能体时代保持自主的方式。
Personal AGI is how you stay under your own power in the age of agents.
所以，从第一天起，就把你的大脑和技能放在你控制的仓库里，在任何一个平台或收购方对它指手画脚之前。
So keep your brain and your skills in a repo you control from day one before any platform or any acquirer has an opinion about it.
斯宾诺莎去世时，他们清点了房间里的物品。
When Spinoza died, they inventoried the room.
两条裤子，七件衬衫，一台镜片磨床，还有一百六十本书。
Two pairs of pants, seven shirts, a lens lathe, a 160 books.
还有锁在书桌里的伦理学。
And the ethics locked in a desk.
他几乎一无所有，但从来没有人能控制他的技能文件。
He owned almost nothing, and nobody ever controlled his skill files.
那个书桌抽屉就是他的仓库。
The desk drawer was his repo.
像他那样拥有你自己的。
Own yours like he owned his.
现在有三个反对意见，我在这里都能听到，所以我们就直接说吧。
Now three objections, and I can hear them from up here, so let's just do them.
第一个反对意见是，模型进步太快了，所有这些辅助工具都会过时。
Objection one, the models are improving so fast that all this harness stuff will be obsolete.
就等下一个版本吧。
Just wait for the next release.
这是那些相信‘更惨痛的教训’的人群，我很喜欢他们，但注意每次模型发布时实际发生的情况。
This is the better bitter lesson crowd, and I love them, but notice what actually happens in every model release.
模型越好，差异化的重点就越转移到上下文上。
The better the models get, the more the differentiator moves to context.
当每个人的引擎都有一千马力时，比赛就赢在驾驶员和地图上。
When everyone's engine is a thousand horsepower, the race is won on the driver and the map.
权重是大家的。
The weights are everyone's.
而知识库是你自己的。
The library is yours.
至少我希望如此。
At least I hope it is.
更好的模型会让你的知识库更有价值，因为更聪明的读者能从同样的书中提取更多内容。
A better model makes your library worth more because a smarter reader extracts more from the same books.
我和这栋楼里的任何人一样，都在为实验室加油。
I'm rooting for the labs as hard as anyone in this building.
但他们每次发布的模型，都是对我已经拥有的员工的一次免费升级，也是我希望你拥有的员工的一次免费升级。
But every release they ship is a free upgrade to a workforce I already own and a workforce I want you to own.
第二个反对意见。
Objection two.
这不就是RAG吗？
Is this just rag?
当然。
Sure.
而且Postgres也就是B树。
And Postgres is just b trees.
检索是基础，不是产品。
Retrieval is the primitive, not the product.
难的是围绕它的一切。
The hard part is everything around it.
首先记录什么，如何丰富和关联，什么提升为热记忆，什么归档为冷参考，当两个事实冲突时由谁仲裁，检索是容易的。
What gets written down in the first place, how it gets enriched and linked, what gets promoted to hot memory versus filed as cold cold reference, who arbitrates when two facts disagree, retrieval is easy.
值得被检索出来，这才是产品。
Being worth retrieving from is the product.
第三个反对意见，也是我最重视的一个。
Objection three, and it's the one that deserves the most respect.
你把整个生活都放进一个系统里，邮件、会议、孩子的日程。
You put your entire life in one system, your email, your meetings, your kid's schedules.
要是泄露了怎么办？
What happens when it leaks?
我的答案和整场演讲的答案一样。
My answer is the same answer as the whole talk.
正因为如此，它才必须是你的。
That's exactly why it has to be yours.
我的大脑运行在我自己的基础设施上，在我自己的代码仓库里，用我自己的密钥。
My brain runs on my own infra in my own repo, under my own keys.
相比之下，默认情况并不是隐私。
Compare that to the default, which is not privacy.
默认情况下，你的生活已经分散在十朵云上，这些云由利益与你不同的公司拥有，除了你之外，任何人都可以搜索到。
The default is your life is already scattered across 10 clouds owned by companies whose incentives are not yours, searchable by everyone except you.
我并没有因为整合我的信息而制造风险。
I didn't create the risk by consolidating my context.
我只是接管了它。
I took custody of it.
托管就是安全模型。
Custody is the security model.
如果你不信任自己保管钥匙，我保证答案不是更信任别人的服务条款。
And if you don't trust yourself to hold the keys, I promise you the answer isn't trusting someone else's terms of service more.
那么，我为什么要把这一切都开源呢？
So why did I open source all of it?
包括工具、大脑架构、技能，整个个人操作系统。
The harness, the brain architecture, the skills, the whole personal operating system.
人们问我这个问题，因为他们似乎觉得这里面肯定有猫腻。
People ask me this because they seem like they they they think there must be a catch.
嗯，答案是因为我可以。
Well, the answer is because I can.
因为对我来说，在YC工作意味着我不需要靠自己的基础设施来赚钱。
Because being at YC for me means I don't have to monetize my own infrastructure.
但“因为我可以”也回答错了问题。
But because I can is also the answer to the wrong question.
真正的问题是，为什么每个人都应该这样做。
The real question is why anyone should.
答案是，我相信强者的工具应该被分享出去。
And the answer is that I believe tools of the powerful should be given away.
每个时代都有一种私有的杠杆技术，是强者拥有而其他人没有的。
Every era has a private technology of leverage, a thing the powerful have and everyone else doesn't.
很长一段时间里，那是读写能力。
For a long time, it was literacy.
后来，是资本。
Then, it was capital.
现在，今天，就是它，这个工具，这个库，这个由Markdown组成的劳动力。
Right now, today, it's this, the harness, the library, the workforce made of markdown.
拥有它的人，正在悄无声息地以不同于没有它的人的规模运作。
The people who have it are quietly operating at a different scale than the people who don't.
而且这个差距每个月都在拉大。
And the gap is widening every month.
这就是整个会议的主题，赋予你力量，让你能够自己做到。
And that's what this whole conference is about, to give you the power to be able to do it for yourself.
当这样强大的东西保持私有时，你会得到一种祭司阶层。
When something like that, that powerful stays private, you get a priesthood.
当它被分享出去时，你会迎来一场文艺复兴。
When it gets given away, you get a renaissance.
我知道我想生活在哪个时代，这意味着我可以做我真正相信的事情。
I know which one I want to live in, which means I get to do the thing that I actually believe in.
我会把它作为信条给你，因为这是我最接近信条的东西。
And I'll give it to you as a creed because it's the closest thing I have to one.
说别人不敢说的话。
Say the things other people won't.
资助别人不敢资助的人。
Fund the people other people won't.
建造别人不敢建造的建筑。
Build the buildings other people won't.
编写并免费发布别人不敢发布的代码。
Write and give away the code that other people won't.
离开别人不敢离开的机构。
Leave behind the institutions that other people won't.
当你公开建设时，你应该知道接下来会发生什么，因为斯宾诺莎的故事还有一章。
And when you build in the open, you should know what's coming because Spinoza's story has one more chapter.
1676年11月，戈特弗里德·莱布尼茨，欧洲最耀眼的奇才，穿着丝袜，行李中带着一台计算器，前往海牙，在一个阁楼里与欧洲大陆最遭人恨的人共度了三天。
November 1676, Gottfried Leibniz, the most glittering genius in Europe, silk stockings, a calculating machine in his luggage, travels to The Hague to spend three days in an attic with the most hated man on the continent.
然后他在接下来的四十年里对此撒谎。
And then he spends the next forty years lying about it.
公开场合，这次拜访只是匆匆几小时。
Publicly, the visit was a few hours in passing.
私下里，他的笔记里满是关于斯宾诺莎的痴迷评论。
Privately, his notes are crammed with obsessive commentary on Spinoza.
我每周都喜欢这种小规模的版本。
I love a small version of this weekly.
我说现在我的大部分代码都是智能体写的，然后嘲讽中午前就来了。
I say agents write most of my code now and the dunks arrive by lunch.
然后我看看那些最响亮的嘲讽者实际在发布什么，结果全是智能体。
Then I look at what the loudest dunkers are actually shipping, and it's agents all the way down.
所以现在就要学会这个模式，因为公开构建保证你会需要它。
So learn the pattern now because building in public guarantees you'll need it.
首先，他们引用转发你，然后他们克隆你。
First, they quote tweet you, then they get clone you.
那些嘲讽只是采用曲线在自我宣告。
The dunks are just the adoption curve announcing itself.
我想让你看看，当这种架构指向唯一真正重要的东西时，它是什么样子的。
And I want to show you what this architecture looks like when it's pointed at the only thing that really matters.
我有个朋友，他的儿子患有一种罕见的癫痫。
I have a friend whose son has a rare form of epilepsy.
没有实验室，没有经费，没有许可。
No lab, no grant, no permission.
他就直接去了，而且你确实可以就这么去做。
He just went, and you can just do things.
他建了一个包含八万个Markdown文件的仓库，相当于为一个小男孩打造了一个大脑，把自己推到了人类对儿子确切病情认知的极限。
He built a repo of 80,000 markdown files, a brain for one small boy, And pushed himself to the absolute edge of what humanity knows about his son's exact condition.
每一次专家门诊、每一篇论文、每一次癫痫发作记录、每一种药物相互作用，都被索引、交叉关联，随时可用。
Every specialist visit, every paper, every seizure log, every drug interaction indexed and cross linked and ready.
这样，当新医生提出一个想法时，他几分钟内就能知道这个想法是否已经尝试过。
So that when a new doctor has an idea, he knows in minutes whether it's already been tried.
一个父亲，一台笔记本电脑，和一个图书馆。
A father, a laptop, and a library.
这就是个人AGI。
That is personal AGI.
不是基准测试，也不是演示。
Not a benchmark, not a demo.
我今晚描述的整个架构——图书馆、图书管理员、在正确时刻打开的三本正确的书——都指向一个人在这个世界上最爱的那件事。
The entire architecture I've described tonight, the library, the librarian, the right three books open at the right moment, aimed at the one thing, one man loves the most in the world.
没有人会来为他建造那个。
Nobody was coming to build that for him.
所以他自己建了。
So he built it.
也没有人会来为你建造你的。
And nobody is coming to build yours for you.
这是好消息。
That's the good news.
你被告知需要的一切——团队、资金、许可、资质——都只是因为一个人无法同时记住七件事并每天工作十六小时而想出的权宜之计。
Everything you were told you needed, the team, the funding, the permission, the credential was a workaround for the fact that one person could hold seven things in their head and work sixteen hours a day.
这个事实刚刚失效了。
That fact just expired.
你现在可以飞了。
You can fly now.
不是比喻，是机械意义上的飞。
Not metaphorically, mechanically.
每一个你曾想‘我要是有这个人就好了’的问题。
Every problem where you thought, I wish I had this person.
‘我希望我能雇到这个人，但我请不到。’
I wish I could hire this person, but I can't get them.
现在你可以了。
You can.
每一个大到读不完的档案，每一个复杂到难以清理的数据集，每一个你被告知不要试图煮沸的海洋。
Every archive too big to read, every dataset too gnarly to clean, every ocean you were told not to boil.
我们现在可以把海洋煮沸了。
We can boil the ocean now.
我有一句座右铭，想留给你们。
I have a sentence I live by and I wanna leave it with you.
一切都是编造的，但你可以自己编。
It's all made up, but you get to make it up.
世界上每一个机构，包括那个在1656年对23岁的人念咒语的，都是由不比你聪明的人编造出来的。
Every institution in the world, the one that read a curse over a 23 year old in 1656, was made up by people no smarter than you.
你和之前每一代创始人的区别在于，他们必须先招募几十个信徒才能开始建设。
The difference between you and every generation of founders before you is that they had to recruit dozens of believers before they could build anything at all.
你只需要一台笔记本电脑，以及你已经拥有的几年个人经历。
You need a laptop and a few years of your own history you're already sitting on.
整个活动大约有7000人。
There are about 7,000 people at this whole event.
7000个坎迪斯。
7,000 canadises.
7000个奋斗者。
7,000 strivings.
在历史的大部分时间里，那些奋斗几乎都没有观众。
For most of history, almost all of that striving never got an audience.
它们因为等待资金、等待人手、等待许可、等待别人先相信而消亡。
It died waiting for funding, waiting for head count, waiting for permission, waiting for someone else to believe first.
我今晚展示的这套机制，是我见过的第一个能让奋斗直接投入工作的技术。
The machinery I showed you tonight is the first technology I've ever seen that lets the striving go straight to work.
一个人，没有中间人，没有许可。
One person, no intermediaries, no permission.
我真的觉得世界还没意识到，7000个拥有这种力量的人走出大楼后会做出什么。
I genuinely do not think the world understands yet what 7,000 people with that kind of leverage walk out of a building and do.
斯宾诺莎在《伦理学》结尾写了九个字，这本书当时不得不藏在书桌里偷偷带出去。
Spinoza closed The Ethics, the book that had to be smuggled out in a desk with nine words.
一切卓越的事物，既困难又稀有。
All things excellent are as difficult as they are rare.
而困难刚刚崩塌了。
The difficulty just collapsed.
稀有性现在就取决于你了。
The rarity is now up to you.
去创造吧。
Go and build.
谢谢。
Thank you.
关于 Bayt 播客
Bayt 提供中文+原文双语音频和字幕，帮助你打破语言障碍，轻松听懂全球优质播客。