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。
如果我们做的只是帮我总结一下这封邮件，那你其实没有真正利用到新智能的优势。
If all we're doing is, like, summarize this email for me, you're not really benefiting from the new intelligence.
我是费利克斯·里森伯格。
This is Felix Rieseberg.
他的团队大概用了一个半星期就搭建出了Claude Cowork。
His team built Claude Cowork in about a week and a half.
现在他负责Anthropic公司Claude的工程，而且他不再写代码了。
Now he leads engineering for Claude at Anthropic, and he doesn't write code anymore.
我只是Claude的一个传送带。
I'm just a conveyor belt for Claude.
Claude需要能够访问所有的崩溃数据，然后自己判断哪些需要修复，哪些不需要修复。
Claude needs to have access to all the crash data and then needs to, like, figure out itself what it needs to fix, what it doesn't need to fix.
那还有什么是你不会放心交给智能代理去做的吗？
Is there still anything you would not trust an agent with?
我真的很希望当前的这个会话能够自我评估。
I really want the current session to judge itself.
我觉得这个还是很不靠谱。
That I think is still really iffy.
我怎么才能防止人工智能自信地向我展示错误信息？
How do I protect myself from AI just confidently presenting mistakes to me?
我们很多人都遇到过这种情况。
It happens to a of us.
我有一个非常实际、简单的答案。
I have like a very practical, easy answer.
菲利克斯，非常感谢你。
Felix, thank you so much.
我很高兴你能成为Claude的用户，还是个热情的用户。
I'm very excited to have you as a user of Claude, like an avid user.
我们这个频道的所有运营都是用Claude来做的。
We're using Claude for all the operations on this channel.
每个团队成员都有自己的项目来处理各自负责的事情。
Every team member has their own project for the things they're responsible for.
作为一个运营者，对于创作者来说，我的团队算是相当大的。
And as someone who runs it, it's pretty big team for a creator.
我有35名员工，我希望他们尽可能高效。
I have 35 people and I want them to be as efficient as possible.
我希望产出尽可能多的内容。
I want to produce as much as possible.
但今年我也注意到，我们成功推出了新的社交媒体渠道。
But also something that I've noticed this year, we were able to launch, I think, new social media channels.
所以我们推出了一个新闻通讯。
So we launched a newsletter.
我们一年半前开通了LinkedIn，因为它是文字为主的。
We launched LinkedIn a year and a half ago because, you know, it's text based.
所以AI在很多方面帮了我大忙。
So AI has enabled me in many, many ways.
而这次对话的目标，就是让我比所有观众和听众都更先进。
And the goal of this conversation is to make me even more advanced than all of the viewers and all the listeners.
从我看到的你做的事情来看，你好像已经挺先进的了。
I mean, from what I've what I've seen you do, it seems like you're pretty advanced as it is.
但是成为...是没有边界的，对吧。
But there is no borderline, right, to to being as
商业赞助？
a business sponsor?
那倒是真的。
That's true.
我觉得我每天都在发现自己这方面的新东西。
Think I'm discovering that about myself every single day.
而且我有点能...
And I can sort of...
在Claude上工作最好玩的事情之一，就是这种‘我造了这个产品’的感觉。跟我的其他产品相比，用那些产品的时候我很清楚你会拿它来干什么。
One of the most fun things of working on Claude by far is this notion of I built this product and in comparison to my other products, with my other products, it was pretty clear to me what you're gonna use this for.
比如我经常举的例子是，我在Slack工作的时候，大概能猜到你会用它来做什么。
Like the example I always go with is I worked on Slack and kinda have a feeling of what are gonna do with this.
做Claude最神奇的地方就是，我看到大家真的能飞起来，做出我没想到能做到的惊人事情。
The thing that's magical about working on Claude is that I see people really fly and do amazing things that I didn't consider possible.
对我来说，作为产品开发者，看到大家用你做的这个东西飞起来，真的是太开心了。
And that's like a that's like for a product builder, that's such a treat, you know, to see people fly with this thing that you made.
你有没有见过特别有野心的事情，然后心想：天哪，我想模仿这个？
Is there something you've seen that's super ambitious and you're like, oh my god I want to copy this?
模仿倒谈不上，不过两天前，有……
I'm not sure about copying, but two days ago there's there's...
我记得他其实是个产品经理，叫帕维尔。他让Claude处理大量NASA的数据，然后让Claude帮他找一颗新行星，结果还真找到了。
I think he's actually a product manager by trade, his name is Pavel, and he pointed Claude at a large amount of NASA data and asked Claude to find him a new planet and it did.
不过目前还没有确认。
So apparently it is not confirmed yet.
NASA正在努力确认这个发现。
NASA is working on actually confirming it.
我在这里有点不太确定，因为我不是什么天体物理学家。
I'm on a little shaky ground here because I'm not an astrophysicist of any kind.
但显然，克劳德可能发现了一颗我们以前不知道的新行星。
But apparently Claude may have found a new planet we didn't know of before.
这太神奇了。
This is fascinating.
好。
Okay.
而且这种事我根本想不到。
And it's the kind of thing that I would have never considered.
但如果你听听这个故事，我觉得对我们这些做知识工作的人来说，把AI指向大量数据，然后就能发现有趣的东西，这本身听起来很合理，我们每天都这么干。
But if you hear the story of I think for us who work maybe on more knowledge work y things, right, the idea of point AI at a large amount of data and you're gonna find interesting things, That in isolation seems pretty reasonable, something we do every single day.
但想到去看看我们关于太空的所有公开数据，把它当成类似的问题，去里面找找有没有我们以前没注意到的东西，这种想法真的很美。
But the idea of looking at all of the public data we have about space and considering it a similar problem of go and see if you find anything in here we haven't noticed before.
太美了。
So beautiful.
你跟我聊聊你的一天吧？
Can you talk to me about your day?
你帮忙建了Claude Code Work，但你说你不再写代码了。
You helped build Claude Code Work, but you said you're not coding anymore.
那早上起来都做什么呢？
So what happens in the morning?
你具体做些什么？
What do you do?
是不是你的智能体在跟你汇报当天的情况，还是怎么回事？
Is it like your agents are briefing you on what is coming or what is it?
对，我的智能体确实会简单跟我汇报一下当天的安排。
That is true, my agents do brief me a little bit on what's coming for the day.
我确实在思考问题的层次上提升了不少，对吧？
I've really levelled up in sort of altitude on the problems that I'm thinking about, right?
有时候我觉得，现在我更关心人的情感、同理心，试图理解人们是怎么运作的，而几年前作为传统的工程经理，我可能没那么关注这些。
Like sometimes I feel like these days I'm much more concerned with human emotion, empathy, trying to understand how people work in a way that I think I was less so as a as a classic engineering manager a couple of years ago.
所以对我来说，我真正发现的是，需要向其他专业人士求助的次数真的减少了。
So I think for me what I'm really discovering is that the amount of times I have to reach out to other professions to get their help has really dwindled down.
我现在是个很奇怪的角色，既是产品经理又是工程经理，还有传统的人事经理和数据科学家的角色。
I'm now this weird mix between, like, a product manager and engineering manager, just a traditional people manager, a data scientist.
具体回答你的问题，大部分时间我都在忙着做各种原型。
To answer your question very specifically, most days I'm I'm busy with a lot of prototyping.
我们会尝试很多东西。
We just try a lot of things.
会花时间去思考一些问题。
We sort of, like, think about stuff.
我们会考虑几天，看看这个想法到底好不好。
We sleep over it for a couple of nights, try to figure out if it's a good idea or a bad idea.
然后我们会让Claude帮我们实现各种不同的变体和想法，接着再进行评审，看看哪些想法是好的，哪些不好。
And then we just go tell Claude to, like, build all the various variations and ideas we had, and then we need to do critique, try to figure out which ones of these have good ideas, which ones are bad.
不过我觉得原型制作才是平时花时间最多的事情。
But I think prototyping is really what I spend most of my day on.
那你如何判断什么是一个好主意呢？
How do you decide what is a good idea?
比如，你都在看哪些信号？
Like, what are the signals you're looking at?
这个问题困扰人们很久了，我也不敢说自己完全知道什么是好主意，什么是坏主意。
This has been difficult for people for a long time now, and I'm not sure that I have all the answers to, like, what is a good idea, what is a bad idea.
我觉得总的来说，我们判断好坏想法的方式，那个衡量标准已经变了。
I think generally speaking, the way we look at good or bad ideas, the calculus has changed.
想想过去十年、十五年我们做的产品，我们总是会隐性地做成本决策。
Like, if you think about products we've built in the past for the last ten, fifteen years, we we always made implicit cost decisions.
我们得搞清楚，尤其是软件，我能不能做出来？
Like, we we sort of had to figure out, okay, especially for software, can I build this?
维护成本是多少？
What is the maintenance cost?
长期成本是多少？
What is the ongoing cost?
我得投入多少时间？
How much how much time do I invest in this?
机会成本是多少？
What is the opportunity cost?
这些都是工程经理要做的决定，决定要不要去开发这个东西，是或否。
All of these decisions that engineering managers make to figure out do I go and build this, yes or no.
我觉得现在我们可以更快地看到最终结果，直接说，好，给我建个项目，让我玩玩看，试试看。
And I think now we can fast forward to the end state much quicker and be like alright, just build me a project, let me play with this, let's try it out.
我觉得人的注意力还是有限的。
I think human attention is still rather limited.
比如说，你吧，对AI很有品味，对吧？
Let's say you, for instance, have a lot of taste in AI, right?
我大概每天能给你展示一个东西，我很容易就能做出一千个，但我没法给你展示一千个。
Like I can probably show you maybe one thing a day, I can easily build a thousand, I can't show you a thousand.
就算你愿意跟我坐一整天，过一会儿你也会看花眼，然后你就会说，行了，我差不多了。
Even if you're willing to sit down with me for all day, like at some point your eyes are gonna glaze over and you're like okay, I'm kinda done with this.
那么，什么才是一个好产品呢？
So what what makes a good product?
我觉得，尤其是在现在这个AI时代，有句老话说得好，完美就是你再也减不掉任何东西的时候，而不是说做一个什么功能都有的产品。
I think, especially now that we live in this AI world, there's this old saying that like perfect is sort of when you can't take anything away anymore, and less so about like building a product that has every single feature under the sun.
我们可以做很多按钮，但我觉得好的产品其实按钮更少。
Like we can make as many buttons as you want us to, but I think good products actually have fewer buttons.
我大概能知道一个产品好不好用，就是把它拿给几个人试，如果他们很快就上手了，而且能做一些以前做不到的事，那就是行。
I kind of know whether or a product really works if we put it in front of a few people and you see things click for them rather quickly, and they can do things that they couldn't do before.
那个……
That...
那才是我真正在意的信号。
That's really the signal I'm looking for.
你怎么看你在解决的根本问题？
How do you think about the underlying problem that you're solving?
比如说，如果你现在来我们公司，你会怎么考虑哪些任务完全交给AI，哪些还是留给人工，或者是只自动化一部分？
So for example, if you came to my company now, how would you think about the tasks that you would fully automate with AI versus just leave it to humans versus maybe just automating parts of it?
我首先想说，我对自动化和增强的看法稍微有点不同。
The first thing I wanna say is that I think about automation and augmentation in slightly different ways.
如果自动化能帮你省去那些阻碍你享受工作的事情，那它真的能赋予你力量。
Automation can be really empowering if you automate away tasks that sort of keep you from doing work that that you enjoy.
或者如果这些事情很昂贵，自动化能帮你省下很多钱或时间。
Or if it's like expensive and automating it away, like, saves you a lot of saves you a lot of money or time.
在这种情况下，自动化是有用的，我认为这可能是AI目前创造最多经济价值的地方。
In those cases, automation is is useful, and I think it's really where AI currently probably provides most of the economic value.
比如，当我们评估某个东西值不值得时，通常都会看自动化带来的好处。
Like, when we evaluate is this worth it or not, we usually look at automation.
不过，我觉得最有趣的领域，也是人们能借助AI做到以前做不到的事情的地方，是增强。
The area I find most interesting, though, and I think where people really do things that weren't possible with AI before, is augmentation.
这个想法是，AI帮你完成那些因为缺乏技能、技巧或基本知识而无法亲手完成的事情。
It's this idea of AI is helping you do things that you couldn't do by hand because maybe you're lacking skill or techniques or just the raw knowledge to do something.
当我去一家公司时，最让我印象深刻的往往不是工具本身做什么，而是你如何改变工作流程。
And when I go to a company, the thing that is often most impressive to me is less so what the tool does and more so how you change the workflow.
我说的有道理吗？
Does that make any sense?
AI是一种有趣的产品，因为要真正利用好它，你得重新思考做事的方式。
AI is an interesting product because to really harness it, you kind of have to rethink how you do things.
你得想想自己的目标是什么。
You have to, like, think about your outcomes.
你得弄清楚什么算好，什么算坏，比如，我怎么判断自己要求的结果是否真的好。
You have to figure out what does good mean, what does bad mean, like, how do I evaluate whether or not this thing that I'm asking for, like, whether the result is good.
通常给我留下最深印象的人，就是那些特别善于反思自己工作流程的人。
And the people who impress me the most are usually the ones who are very good at coming up with, like, good at reflecting on their own workflows.
但这太难了。
But it's so hard.
比如我思考自己工作的时候，我知道很多人用AI来起草文本。
When I'm thinking about my work, for example, I know a lot of people use AI for drafting texts.
比如，怎么
Like, how
定义一篇文章好不好呢？
do you define a good text?
比如说，对于知识型工作，这还可能吗？
Like, is it still possible for, like, knowledge work?
因为对于编程，我觉得是可能的，对吧，因为你知道你想要什么，你不想要任何bug。
Because for coding, I guess it's possible, right, because you know what you want, you don't want any bugs.
但是当涉及到知识型工作，以及人类创造力的成果时，你怎么定义什么对智能体来说是好的呢？
But when it comes to knowledge work and, like, the result of human creativity, how do you even define what's good for the agent?
这真是个棘手的问题。
It's such a tricky problem.
比如，我经常举的例子是，假设你的工作是报道天气。
Like, the the example I often go with is, let's say your job is to, like, report on the weather.
然后你说，智能体，请查一下外面的温度是多少。
And you're like, agent, please figure out what the temperature is outside.
这有两种看法，因为你需要重新思考什么是好的。
There's, like, two ways to look at that because you have to rethink what good means.
如果模型只是猜对了温度，那不应该算及格。
Like, if the model just guesses the temperature and is correct, that should not be a passing grade.
那应该是很糟糕的。
That should be bad.
我们应该重新思考我们在这里做的事情。
We should rethink what we're doing here.
但如果模型实际上偏差了半度左右，但它真的找到了正确的位置，找到了一些可靠的数据来源，而且它获取结果的方式也是可靠和好的，那实际上应该比一个碰巧猜对的好结果更好。没错。
But if the model is actually off by half a degree or something, but it really figured out the right location, figured out some reputable data sources, and sort of the the the way it got there was reputable and good, that should actually be better than like a gassed good result Mhmm.
如果这说得通的话。
If that makes any sense.
但对于写作和其他所有你认为涉及很多品味的输出，这些结果的好坏有点难以证明或验证，很难说一个结果比另一个好。
But for writing and like all the other outputs where you feel like a lot of taste is involved and it's like a little hard to like prove or verify whether or not the result is better than another.
这确实棘手，但我觉得我很喜欢和那些写得很好而且对AI有点友好的人聊天。
It is tricky, but I think I really enjoyed talking to very good writers who are also a little AI friendly.
他们怎么说？
What do they say?
看他们工作真是令人着迷。
It is it is fascinating to watch them work.
我的朋友亚当·戴维森，他在好几家不错的媒体当过撰稿人，他给Claude准备了一堆各式各样的提示词，比如说，把我的论点放大一百倍，稍微改改我的文字，你最不同意哪些地方，最有力的反驳是什么。
My friend Adam Davidson, who was a staff writer at a number of pretty good publications, he has all of these various prompts for Claude where he's like, okay, take my argument and magnify it by a 100, just mess with my text a little bit, what are the things you disagree with the most, like what is the strongest counterargument.
有意思的是，对我来说，作为一个不是作家的人，我觉得最直接的做法就是给Claude一段文字，然后说，把它改好。
It's interesting because I think to me as someone who's not a writer, I think the naive approach would be given a text Claude, make this good.
但我越来越发现，在知识工作里，秘诀其实不是要一个更好的结果，而是在中间多做一步，就是看看这个领域里最厉害的人是怎么做的。
And I think what I'm discovering more and more is that for knowledge work, the secret sauce is like less than asking for a good output and sort of doing one step in between, which is the best in the field, how do they work?
我们能不能把这些步骤自动化呢？
And can we automate those steps?
这可能能模仿他们的工作流程，但品味还是缺不了。
This would kinda mimic their workflow, but still the taste isn't there.
这就是为什么我觉得你还是需要很多
That's why I think you still need a lot of
我觉得这话没错。
I think that's true.
我觉得这话没错。
I think that's true.
但有趣的是，如果你跟真正优秀的作家聊，很多人不会认为文章好是因为他们天生有那种点石成金的本事。
But the interesting thing is that if you talk to really good writers, most of them would disagree that the text is really good just because they have, like, the magic touch.
他们常常会说，不不不，这是努力的结果。
They would often say, no, no, no, this is the result of hard work.
有意思。
Interesting.
就像，大多数艺术家都有自己的创作流程。
Like it's this idea of, like, most artists have a process
而且他们中的很多人可能会觉得……
And many of them would actually probably find it...
我不想用‘冒犯’这个词，但如果你……
I don't wanna say insulting, but, like, if you...
我越跟艺术家聊，越思考这个问题：你的作品好，是因为它对你来说很自然，随随便便就能写出好东西吗？
The more I talk to artists and the more I explore this idea of, like, is your stuff just good because it comes natural to you and it just happens to be good?
还是说你有一套多年来不断打磨的流程？
Or do you have a process that you refined over the years?
很多人对自己的创作流程相当自豪。
A lot of them are quite proud of their process.
我就会说，这真的是非常非常辛苦的工作。
I'm like, this is very, very hard work.
就是说，要把它做好，需要很多一步一步的细致操作。
Like, there's a lot of manual individual steps going to making this good.
并不是说我一坐下来，突然就冒出一本好书来。
It's not like I sit down and suddenly there's a book that's great.
如果你去作家的工作室，对吧，那里有很多白板，上面写满了各种信息。
And if you go to writer studios, right, there's a lot of whiteboards and a lot of information.
而且我在很多不同的领域都注意到这一点：Claude 很愿意接受这样的指令——不是说我想要一个看起来差不多的结果，而是说，Claude，我希望你像这个领域的专业人士那样，采取同样的步骤。
And I'm noticing this across a number of different domains where Claude is quite receptive to being told not I want a result that looks like, uh-huh, but Claude, want you to take the same steps that a professional in this field would take.
这个比喻用得比较多。
This is like very high mileage.
让我稍微简化一下。
Let me bring this down a little bit.
我觉得现在大家能在公开场合看到很多关于动态、视频和三维的东西，我觉得Opus 5.5进步特别大。
I think we're currently in the public seeing this quite a bit with like motion, video, and three d, where I think Opus 5.5 has made a pretty big jump.
但是大家最大的突破就是直接告诉Claude用Blender。
But the big jump that the public has made is to just tell Claude to use Blender.
就是说，Claude，请用专业动画师用的那些工具。
Just like, Claude, please use the same tools that professional animators use.
结果现在效果还挺好的。
And now the results are quite good.
真正的解锁点在于媒体、技术和工具，而不是在于对什么才是一部好的动画电影的原始理解。
The unlock really is here in the media and the technique and the tools and in this less so like, raw understanding of what makes a good animated movie.
如果你在Spotify上收听，而且这次的对话对你有用，请一定给我们节目评分。
If you're listening on Spotify and this conversation was useful, please, please, please rate our show.
我看到了你们的浏览量，但是没有收到实际的反馈。
I see all your views, but I'm not getting the actual feedback.
这真的能帮我们团队决定下次拍什么，也能让从未听说过我们的人决定要不要来听一听。
This can really help our team to decide what to film next, and it can help someone who's never heard of us decide whether to give us a listen.
非常感谢。
Thank you so much.
但我在想，对于普通人来说……
But I wonder for people...
你提到了这个提示，就是假装你是这个领域的专家。
Like, you mentioned this prompt, act as if you're a professional in this field.
然后它会去网上搜索，但很多信息其实根本不在那里。
And it's gonna go on the Internet and research, a lot of information is just not there.
你怎么相信它给出的结果呢？
How do you trust the output?
比如说，如果我想开始做另一个媒介，或者去发现星星。
Like, for example, if I want to start another medium or like do discover stars.
我对那个完全不了解，但我只是告诉它，表现得像个资产研究员。
And I have no idea about that, but I just tell it, act like an asset researcher.
我该怎么保护自己，不让AI自信地给我呈现错误信息呢？
How do I protect myself from AI just confidently presenting mistakes to me?
这倒是个好问题。
That's a good point.
我觉得这种情况其实很多人都会遇到。
I think it still happens to a lot of us.
我有个很实际、也很简单的解决办法，就是对抗性审查。这个方法在编程、数学这些领域已经被证明挺有效的，而且我们还能用数字来告诉你它有多管用。
I I have I have, like, a very practical easy answer, which is adversarial reviews, which has proven quite effective in coding, math, and like various other areas where we can give you numerical results on like how effective it is.
对抗性审查说白了就是，你给一个Claude分配任务，再另外开十个，让它们想办法把这个任务拆解、攻击。
Adversarial reviews basically boil down to the idea of you have given given one Claude a task, spin off another 10 net, try to, like, break this down.
它们的主要工作就是找茬，指出哪里不对，证明给你的方案为什么是错的。
Their whole task is to attack this and prove to you why it's wrong.
这跟红队测试或者做空的概念其实是一回事。
It's essentially the same concept as red teaming or short selling.
你是用同一个模型来做这个，还是另外用一个模型？
Are you using the same model for that, or you want to use another model?
这个要看任务的情况。
Kinda depends on the task.
这要看具体情境，但通常对抗性评审……
This is a little context dependent, but usually the adversarial review...
有时候找出问题比创造新东西更容易，人也是这样。
It is it is sometimes easier to find flaws than to create new things, much like as it is with humans.
对我来说，说新手机好不好要容易得多。
It's much easier for me to say whether or not the new phone is good.
这比从零开始设计一部新手机要容易多了。
It's much easier for me to do that than to design a new phone from scratch.
所以通常来说，你可以用智力稍低一点的模型，也能得到很好的结果。
So usually, you can, like, take a step down in intelligence and still get very good results.
这就是非常实用的建议。
That's the very practical advice.
那你会怎么
How would
给它写提示词呢？
you prompt that?
如果已经有一个流程了，那你会添加什么额外的提示词呢？
Like, if there is a process already, what's the additional prompt that you add?
那你先得有个输出结果。
So you have some kind of output.
比如，假设它是一份幻灯片或文档。
Like, let's say it's a slide deck or a document.
或者对我来说，它可能是……
Or in my case, it's...
你提到我今天早上写的那篇博客文章。
You talked about this blog post that I wrote this morning.
我就直接告诉Claude，非常……
I just tell Claude very...
我正在使用Claude Projects，这是我们新出的产品，可以让你在多线程上更轻松地工作。
I'm I'm using Claude Projects, which is this new product that we have that allows you to do work on multiple threads with a little bit more ease.
我基本上就是告诉Claude，帮我开五个线程，让每个线程都去验证，比如，这个帖子哪里错了，我是不是想错了，事实依据是不是有问题，或者我是不是误解了什么。
And I basically tell Claude, please fire off five threads and task each of them with, like, proving that this this post is wrong, that I'm thinking about it the wrong way, that I have my facts wrong, that I misunderstood something.
然后它们就开始执行任务，目标就是拆解我的观点，证明我是错的。
And then they go off with a given goal of, like, break this down, you know, prove that I'm wrong.
那你如何处理所有这些输出内容呢？
How do you then deal with all the output?
因为我也这么做，但我在准备面试时会用不同的模型。
Because I I do the same, but I use I just use different models when I'm prepping for the interview.
比如，告诉我我是不是在这个那个方面钻研得太深了。
Like, tell me if I'm digging too deep into this or that.
但然后我收到了这么多不同的反馈，我就想，好吧。
But then I'm giving so much different feedback, and I'm like, okay.
现在我得做决定，因为这个模型给了我这些，那个给了我其他的，还有一个给了……
Now I have to decide because this model gave me this, this gave me something else, and gave...
这个给了我完全不同的答案。
This gave me something completely different.
你怎么
How do you
那你怎么决定用哪个呢？
then decide what to use?
我觉得你会喜欢这个，因为你大概能猜到接下来要讲什么了。
I I think you're gonna enjoy this because you can see where this is going.
答案还是——Claude。
The answer is once again Claude.
我们在编码上下文里用的是我们称之为“法官”的东西。对。
What we use in encoding context is something we call a judge Mhmm.
你从那些人的输出中获取结果……
Where you get the output from the people who are...
就是那些负责攻击你输出的智能体的输出。
Like, the agents whose job it was to attack your output.
他们会说，好的，这是我们主要的批评点。
They're like, alright, here's our main points of criticisms.
然后你来决定。
Then you Decide.
再用另一个 Claude，他们的任务是……
Use yet another Claude and their job is...
你的任务是验证、核实，并且判断这些反馈是否有效。
Your job is to validate and verify and opinionate whether or not this is valid feedback.
然后，有时候你可以做的，就是最终根据你工作流程的设置，把有效的反馈送回给真正生成内容的那个 Claude，而那个 Claude，也就是 agent，也可以自己判断要不要采纳这些反馈。
And then what you can sometimes do is that ultimately, depending on how you set up your workflow, the feedback that is valid should probably go back to the Claude that actually made the thing, and that Claude, agent, too, can be discerning in whether or not to act on feedback.
因为有时候，我们经常在编程环境下看到，我在知识工作的输出中也看到，有些反馈虽然看起来合理，但可能没抓住重点，或者其实不太可操作，或者没什么用，这通常就是我判断我的输出已经不错的时候。
Because sometimes, and we see this in a coding context fairly often, also see it in, like, the knowledge work output that I make, sometimes the feedback is, like, maybe reasonable but has missed a point or is actually not that actionable or is not that useful, and that's usually when I know that my output is pretty good.
这是一种方法。
That's one way.
那是在你自己都不确定输出质量，而且不太好验证的时候。
That's when you are really quite uncertain whether or not the output is good yourself and you can't easily check.
但另一方面，对于我用 AI 创造的很多东西，其实我不需要完全理解整个过程。
But for a lot of a lot of things that I create with AI on the other hand, actually don't need to fully understand the entire flow.
因为就像我说的，对我个人来说，它……
Because as I mentioned, like, to me as a person, it is...
我几乎每天用的每样东西，我都有自己的看法。
I do have opinions about almost anything I use every single day.
我用这些3D素材重新做了我的网站。
I've remade my website with all of these three d assets.
我根本不会用Blender。
I have no idea how to use Blender.
我从来没用过它。
I've never used it.
但我能大致判断这个东西好不好看，感觉对不对，这也许就够了。
But I can sort of judge whether or not this looks good, feels good, and that might be enough.
就是说，你更像是把自己放在客户的位置上。
Like, you sort of put yourself much more in a client position.
生活中很多事情，我们其实没有特别好的办法去验证结果好不好，除了给提供服务的人反馈。
For so many other things in our life, we don't really have a good way of verifying whether or not we get a good output other than, like, giving feedback to the person we're using.
比如承包商、律师，或者对很多人来说，医生。
Like, for a contractor or a lawyer or, like, for many people, a doctor.
所以你基本上就是把它分成两类。
So you sort of just put it in one of these two buckets.
要么你完全不知道什么算好…
Like, either you really have no idea what good's like...
好是什么样子。
Good looks like.
这种情况下，我的答案就是多用点AI。
There, my answer is more AI.
就是多花点token去试。
Just throw more tokens at it.
要么你心里有数，知道什么算好，那可能就够了，你也不需要非得对过程本身有那么多意见。
Or you do have an understanding of what good looks like and then maybe that's enough and you don't really need to, like, have opinions about the process itself.
我觉得关键是要选对项目，只挑那些你想了解每一步细节的顶级项目，而不是想什么都掌控。
I think it comes down to making sure you select only the top projects where you want to understand every single step instead of just want to control everything.
这就像老问题一样，你得愿意放手，对吧？
It's like the old problem of, like, sort of being willing to let go, right?
没错。
Exactly.
怎么才能不微观管理。
How to not micromanage.
这真的很难。
It's really hard.
现在我们跟智能体也有这个问题。
Now we have this problem with agents.
对于智能体，你有没有什么已经掌握得很熟的经验？
Is there any knowledge for us that you have nailed with agents?
就是你完全信任它们的那种。
Like, you completely trust them.
它们就直接去做了，我也不知道，可能就像——
They just go, I don't know, maybe it's
事实核查。
Fact checking.
对，就像编辑那种事实核查。
Replying Like editorial fact checking.
我现在完全信任它给我的结果。如果让Claude去核对我在幻灯片和这次演讲里说的所有事实，它甚至不需要那么多轮。
I fully trust now the output I get a 100 If I ask Claude to verify that all the facts that I'm stating in the slide deck and this presentation, it doesn't even need that many rounds.
又不是说我用了五十个agent去做这事。
Like, it's not like I have 50 agents doing that.
真的就信任一个agent。
Really trust one agent.
我就说，好。
I'm like, alright.
我需要你为我提到的每个数字找到一手来源。
I need you to find primary sources for every number I mention.
我要你确认这是可靠的一手来源。
I want you to verify that that is a good primary source.
我百分之百完全信任它。
I fully trust it a 100.
你说代理（agent），你们是专门训练了一个负责事实核查的代理，还是就用一个代理处理所有事情？
When you say agent, do you have a special agent for fact checking that you've trained or it's just one agent that you're using for everything?
就是直接用Claude。
It's just just Claude.
我觉得这个观点可能会有点争议，但你应该会喜欢。
I feel like this is going to be a spicy take, which I think you'll like.
我觉得我们在“代理”这件事上有点过度投入了，我可以稍微解释一下。
I think we're kind of over rotating on this agents thing, and I can back this up with a little bit.
因为我觉得模型其实是很厉害的多面手，我们已经一次又一次地看到这一点。
Like, I think models are really good generalists, and we've seen this now time and time again.
比如，当我们刚开始在数学领域探索的时候，我觉得第一个真正认真对待数学的公司是谷歌DeepMind，他们训练了一个定制的模型和一套定制的工具，然后用在数学奥林匹克竞赛上，拿了银牌，那时候这是最先进的水平。
Like, when we first started looking at math as an industry, I think the the first company to really look at math like with with a very serious lens was Google DeepMind, they trained a custom model and a custom harness, and they unleashed it on the Marth Olympiad and they got a silver medal, which at the time was state of the art.
但是四五个月之后，所有主要的通用模型都出来了，并且拿到了金牌。
And then four or five months later, all the major generalist models come along and win a gold medal.
我理解为什么人们会使用代理（agent）。
I understand why people use agents.
我理解，因为如果给AI戴上一顶小帽子，比如叫它数据科学家，我们人类就更容易理解它能做什么。
I have empathy because it's easier for us humans to rationalize what AI can do if we sort of like put on a little hat, like you're a data scientist.
但那位数据科学家和顶尖的生物学家或数学家之类的，拥有同样的智慧。
But that same data scientist has all the intelligence as like a field leading biologist or mathematician or what have you.
而且我觉得如果我们强迫自己先划分...
And I think we're we're sometimes limiting ourselves in our use of AI if we sort of force ourselves to first cut...
就像，编造出一堆角色来。
Like, make up this cast of characters.
所以我刚才提到的很多事情，我并没有真正让那些代理知道自己的角色。
So for a lot of these things that I just meant, I didn't really have any agents who are being told who they are.
但数据重要吗？
But does the data matter?
因为每个代理，你都会上传特定的数据集。
Because for every agent, you upload certain dataset.
或者说，只要你的Claude能访问你所有的工具，那么每个特定的代理接收到什么数据其实并不重要。
Or just because your Claude has access to all your tools, it doesn't really matter what you feed to each particular agent.
但如果要回答每一个具体问题，它得考虑的数据是不是太多了？
But then isn't it a lot of data for it to consider when answering each particular question?
确实是这样，它能很精准地判断自己该用哪种数据，对吧？
That is true, it's quite discerning in what kind of data it wants to use, right?
我觉得正确的做法不是把数据切成小块，然后给每个代理它想要的那一块，因为很多人如果真的想把工作做好，其实需要看到整个业务的全局。
Like I think the right approach here is not to make these slices of data and then give every single agent the slice of data that it wants because many of us, if we wanna be quite effective at our job, we do actually need the full view on the entire business.
如果只给我们一小部分数据，我们很难高效地工作。
We're not quite effective if we only get a small slice of data.
比如说——
Like the...
作为一个工程经理，如果你只给我代码权限，我可能就没那么大用处了；但如果我还能看到客户数据、听客户通话记录，那就不一样了。
As an engineering manager, if you only give me code access, I'm probably gonna be much less useful to you as opposed to if I also have customer data, access to customer calls.
有时候我真的想听听客户电话，比如我们在推销东西的时候。
Sometimes I do wanna sit in like a like a customer call, like when we're trying to sell something.
我想掌握深度使用信息。
I wanna have deep usage information.
直到你获得关于什么有效、什么无效的业务数据，并且你希望你的代理也能拥有同样范围的数据。
Until you get business data about what is working, what isn't, and you want your agents to have the same range of data.
那么，你如何处理数据量可能很大的问题呢？
Now how do you deal with the fact that it's probably a lot of data?
工具。
Tools.
你给他们工具，并详细记录每个工具背后有什么样的数据，这样他们就不必全部消化。
You give them tools and you document really well what kind of data is behind this tool so that they don't actually have to ingest the whole thing.
你就说，如果你想要业务数据，这里有获取业务数据的工具。
You're just like, if you do want business data, here's the tool to get business data.
所以，当你设置这个的时候……
So when you're setting this up...
因为我在想，你提到这一点时，我是不是应该把那些能访问所有具体数据的项目都迁移掉？
Because I'm thinking, like, when you mentioned this, should I move away from all my projects that have access to all the, like, specific data?
比如说，我不知道。
For example, I don't know.
有一个项目是给嘉宾排名的，里面有完整的流程描述，还有所有旧的脚本，这样我们就知道每个嘉宾的表现如何。
There's a project for ranking a guest, and we have the whole description of the process and then all the old scripts there so that we know how each guest performed.
但现在你这么说，是不是意味着我应该离开那个项目？
But now that you're saying this, does that mean that I should move away from that project?
告诉我的Claude：好的，这个文件夹里装着所有脚本。
Tell my Claude, Okay, we have this is a folder with all the scripts.
如果你想在Notion里查看，我们有每个播客嘉宾表现的详细数据，这就是我们邀请嘉宾的方式。
If you want to look at them in Notion, we have all the intelligence on how every podcast guest performed, and this is how we're in guests.
就把它放在那里。
And just leave it there.
每当我想给嘉宾排名时，就开一个新聊天，什么都不告诉它，直接让它给嘉宾排名，它就会...
And whenever I want to rank a guest, I just start a new chat without telling it anything and just tell it rank the guest, and it's gonna
我对工作流程的规范很谨慎。
I'm careful to be prescriptive in workflows.
我总说，如果对你有效，那对你就是有效的，我不认为你应该改变它。
Like I always say, if it works for you, it works for you, and I don't think you should change it.
问题在于瓶子越来越好了，我也想变得更好。
The problem is the bottles are getting better, and I want to get better.
所以这个方法今年一直有效，但一年时间可不短。
So this has been working for this year, but a year is a lot.
好的。
Okay.
嗯，如果你要问我推荐怎么做，我的建议是——每个人当然都应该做让自己开心、舒服、有动力的事，但确实，现在的模型在工具调用和识别工具调用方面进步了很多。
Well, I mean, if if looking for the recommendation that I give for like what do I do, again, everyone should do whatever makes them happy and comfortable and what makes them powerful, but it is true that the models have gotten so much better at tool calling and discerning tool calling.
工具调用就是模型知道什么时候该用工具、怎么用工具，不会卡在工具调用循环里，也不会误以为找到了数据就一个劲调用工具直到没数据为止。
And tool calling is this process of like the model knows when to use a tool, how to use a tool, doesn't like get stuck in tool call loops, doesn't accidentally say oh I found a data tool call, let me call this tool until there's no more data left.
现在的模型在高效、有效地使用工具方面特别擅长。
The models have gotten particularly good at efficiently and effectively using tools.
所以因为这个原因，我不再事先确定你需要什么样的数据了。
And for that reason, I don't really make this a priori decision anymore on what kind of data do you need.
这也让我能同时开更多线程和会话，而不必依赖固定的流程。
It also allows me to, like, have more threads going and more sessions going without having to rely process.
过程就简单多了。
Much easier process.
你还在用技能吗？
Do you still use skills?
我们现在确实还在用不少技能，不过Anthropic官方的指导以及我个人的建议都是，用技能的时候要特别小心。
We do use skills quite a bit still, although our official Anthropic guidance, as well as my personal guidance, is be really careful with skills.
我们经常给模型灌输一些其实只适用于上一代模型的东西，在AI领域时间过得实在太快了，一个技能的有效期其实很短。
We often tell models things that were sort of relevant for, like, the last generation of models, which time is so compressed in AI, right, that the half life of a good skill is very short.
有时候我们发现，如果把技能完全删掉，模型的表现反而更好。
And sometimes what we see is if you completely remove the skill, you actually get better output.
我记得Bruce Cherni说过，不如就干脆别想了。
I think Bruce Cherni said that, like, just forget about it.
直接写提示词就好。
Just just prompt.
直接写提示词，看看结果怎么样。
Just prompt and see what happens.
确实非常有效。
It is it is very effective.
这显然对那些有某种评估标准、知道什么算好什么算坏的人最有用。
This is obviously most useful to people who have some kind of eval and have an idea of what good and bad looks like.
但我在用技能还是不用技能的时候，都会非常谨慎和挑剔。
But I'm pretty careful and very discerning in when when I use skills and when I don't.
好的。
Okay.
你知道吗，有一个很好的方式来思考这个问题。
You know you know what a what a great way to think about this is?
我喜欢把重点放在模型体验上，而不是用户体验上。
I like to consider the model experience as opposed to, like, user experience.
所谓模型体验，就是假装你暂时是那个模型。
And model experience is you pretend for a second that you're the model.
你就是克劳德。
You're Claude.
如果你完成了交给你的任务，你希望获得什么样的资源？
What kind of things would you want if you had fulfilled whatever task you're being given?
你很快就会发现，你确实需要访问原始数据，需要能使用电脑，可能还需要安装一些软件。
And you very quickly discover that you actually do want access to the raw data, you want access to a computer, you probably wanna be able to install some software.
我希望自己能做些决定，而不是别人已经带着想法来告诉我。
I want to make some decisions myself without somebody telling me what they think already.
我就想直接看到你说了什么，原始数据，然后自己做决定。
I just want to see exactly what you said, raw data, and make my own decisions.
我觉得前期投入会高一些。
Like, I think the upfront cost is higher.
比如，如果你……
Like, if you...
比如，一年半以前，我还在那种状态里，跟Claude说，又发现了一个崩溃。
Like, even a year and a half ago, I was in this world of, like, telling Claude, found another crash.
请你去修复这个崩溃。
Please go and fix the crash.
但我很快就意识到，我不过是给Claude当传送带的。
But it very quickly dawned on me, I'm just a conveyor belt for Claude.
就是去看看我们的崩溃报告，拿一个过来，放到Claude面前。
Like, go look at our crashes, pick one up, move it over, drop it off in front of Claude.
然后我就说，请修一下这个。
I'm like, please fix this.
我就说，好的。
I'm like, okay.
这，这……
This is this is...
不。
No.
我干脆……
I'm just gonna...
Claude需要直接访问所有崩溃数据，然后自己判断哪些需要修，哪些不需要修。
Claude needs to have access to all the crash data, and then needs to, like, figure out itself what it needs to fix, what it doesn't need to fix.
展开剩余字幕（还有 425 条）
前期投入确实很值得，但我觉得也要考虑到确实有这笔前期投入。
The upfront cost is really worth it, but it's also, like, important to, I think, to consider that there is upfront cost.
不过一旦你把这套东西搭建起来，它真的会为你打开很多扇门，我觉得这才是AI带来真正价值的地方。
But once you set this up, it it opens up so many doors for you, and that's where I find a lot of, like, real hard value coming out of AI.
如果你听说过AI智能体，想知道怎么实际搭建一个，我整理了一份指南，大约二十分钟就能带你完成第一个。
If you've been hearing about AI agents and wondering how do you actually set up one, I put together a guide that walks you through your first one in about twenty minutes.
它会帮你选择要自动化什么，而且不需要编程就能跑起来。
It helps you choose what to automate and get it working with no coding required.
你可以通过我的免费通讯订阅获取。
You can get it through my free newsletter.
链接在下面。
The link is below.
对于一个正在听这段对话的人来说，如果他们认为要正确使用AI需要大量准备工作，你会怎么建议？如果某人今天刚开始用AI做知识工作，不需要那些语气之类的东西，比如我的商业策略。
For someone who just who's listening to this conversation and who thought they would need like a major setup to start using AI correctly, would you say if somebody is starting just today, they're using it for knowledge work.
他们不需要那些语气之类的东西，比如我的商业策略。
They don't need all these tone of voice things like my business strategy.
你只要让它访问你的电子邮件。
You just give it access to your email.
它会从中学习。
It learns from it.
从你的语气中，你让它访问你的通讯录或者其他什么。
From your tone of voice, you give access to your newsletters or whatever.
所以你已经不需要那些设置了。
So you don't need any of that setup anymore.
我觉得，今天开始用Claude最好的办法，第一步就是下载桌面应用。
The best way to get started with Claude today, think, is step one, download the desktop app.
它内置了一个浏览器，你可以让Claude登录到你正在使用的某个系统。
It has a built in browser where you can log Claude into some system you're using.
不用管连接器，不用管下载所有数据，不用管各种乱七八糟的东西。
And forget about connectors, forget about, like, downloading all the data, forget about, like, all kinds of things.
就让Claude稍微用一下你的电脑，差不多就是把键盘交给它。
Just let Claude use your computer a little bit, sort of give away the keyboard.
这样你就能走得很远，真正看出哪些有效，哪些没用。
That gets you so far and really shows you what works and what doesn't work.
然后到某个时候你可能会发现，哎呀，Claude 真的在你用的工具上帮了大忙，不管是内部的还是外部的工具。
And then at some point you might discover, oh, Claude is really helping me with whatever tool you're using, whatever internal or external tool you're using.
一旦你觉得某个流程真的很顺手，你可能会想，好吧，能不能给Claude提供更好的方式，而不是手动在这个面板里点来点去？
Once you see that a workflow really works for you, you might consider, okay, can we give Claude something better than, like, having to manually click in this dashboard around?
能不能给它弄个连接器？
Can we give it some kind of connector?
但一开始用浏览器就能让你走得很远了。
But just starting with the browser gets you so far.
好的。
Okay.
然后当它——
And when it...
最后咱们来总结一下这些智能体。
Just to wrap up with the agents.
你说到代理的时候，不只是聊天而已。
When you say agents, that don't...
你基本上就是聊天。
That means you have just basically chats.
就这些。
And that's it.
没有专门做研究的代理，也没有专门报税的代理？
There there is no, like, specific agent who does research, no specific agent who does your tax returns?
没有。
No.
没有。
No.
我的意思是我有项目，所以我会考虑不同的工作任务，但我不会考虑不同的角色。
I I mean, I do have projects, so I do think about I do think about different work items, but I don't think about different roles.
所以我不会……
So I don't...
我从来不会对Claude说，你是顶尖的什么什么。
I I never tell Claude, you're an excellent whatever.
我觉得这招在2024年那会儿特别流行。
I think this was this was very en vogue, sort of like 2024.
我们那时候会说，Claude，你是个超棒的工程师。
We were like, Claude, you're an excellent engineer.
请帮我把这里的问题都搞定。
Please go fix whatever problems are here.
但你现在的提示词里还在用这招吧。
But you still use that in prompting now.
一开始我们就是用这种“假装你是某某专家”的方式。
Act like something that we started with.
假装你是顶级专家什么的。
Act like a top whatever.
我特别想让Claude告诉我，这个领域的顶尖高手会用哪些工具。
I specifically implore Claude to, like, tell me what tools some on top of the field in this field would use.
然后我就说，我希望你使用同样的领域。
And then I'm like, I want you to use the same fields.
现在大多数模型仍然存在的一个问题是，如果你只是简单地告诉它……
A problem most models still have today is that if you simply tell it...
数据科学也许是个不错的例子，或者像会计。
Data science is maybe a good example, or like accounting.
很多模型仍然有这样的问题：如果你让它们扮演一个非常优秀的会计，它们会进行一些角色扮演，而角色扮演并不是我们真正想要的。
A lot of the models still have the problem that if you tell them to act like a very good accountant, that they engage in a little bit of role play, and role play is not really what we're going for.
我们想要的是专业人士会使用的技巧、工具和知识。
We want the techniques, the tools, the knowledge that a professional would use.
我们并不真的只想要那些术语。
We don't really want just the vocabulary.
比如说，我不需要Claude听起来像……
Like, don't I don't need Claude to sound like...
你不需要穿上戏服。
You need to be costume.
其实我想让它像跟五岁小孩说话那样跟我聊，但同时用上这个领域里所有顶尖的工具。
I actually want it to talk to me like I'm five, but I wanted to use all the top tools in the field.
那有没有什么事是你还是不愿意交给智能体去做的？
Is there still anything you would not trust an agent with?
这个问题不好回答。
It's a tricky question.
我们可以一层一层地往上加更多验证。
We can just keep layering more validation on top.
所以对我来说，问题更在于……
So I think the question for me is much more...
与其说我不信任Claude做什么，不如说我不信任的是单次会话能搞定哪些事。
It's less about what do I not trust Claude with and much more about what things do I not trust one single session with.
而且我觉得，我不放心单次会话搞定的清单，其实比我不放心交给Claude做的清单要长得多。
And I think the the list of things where I don't trust one single session is actually much larger than the list of like things I wouldn't trust Claude with.
我很少想让当前这个会话自己来评判自己。
I rarely want the current session to judge itself.
这个我觉得还是很不靠谱。
That I think is still really iffy.
比如说，当前正在运行的Claude会话，不管是在Claude Code、Claude Cowork、聊天里，还是在项目里，它对于自己做的活儿到底好不好，往往判断得很差。
Like the current Claude session that is running, wherever you have it, in Claude Code, in Claude Cowork, in Chat, and even in a project, the current session tends to be quite bad at being a being a judge of whether or not the work it did was quite good.
所以，我尽可能去找外部的信号。
So as much as I can, try to find that outside signal.
如果实在没办法，外部信号可以是另一个Claude会话。
And the outside signal in a pinch can be another Claude session.
这个方法已经被证明相当有效。
That has proven to be quite effective.
但我更倾向于去找一些硬性的外部证据，来判断什么有效什么没效。
But my my preference is to find some hard outside evidence for what works and what doesn't work.
而且有时候你得稍微信任一下过程，给Claude一点时间。
And sometimes you need to like trust the process a little bit and give Claude a little bit of time.
比如，在你们这个领域，我随便举个例子啊，我猜也许用户数或者订阅数是个重要的指标，这个……
Like, for instance, in your field, kinda making this up, but I assume that maybe, like, the number of users or subscribers is, like, an important metric that
哦，是浏览量。
you Oh, the number of views.
这就是我们……
That's that's what we're...
每次我给Claude布置任务的时候，我会告诉它，我们需要达到这个浏览量。
Whenever I give a task to Claude, I tell her, like, we need to hit this number of views.
去试试吧。
Go and try.
我们还在努力解决的另一个问题就是这个反馈循环，对吧？
And that's something that we're still struggling with is this feedback loop, right?
因为是我给我的团队反馈，然后我的团队实际上发布的是别的内容，不是Claude给我们的。
Because it's me giving feedback to my team and then my team actually going and posting something else, not what Claude gave us.
大概有70%是一样的，但我加的那30%其实是我个人的品味。
It's like maybe 70% the same, but then the 30% that I've added is actually my taste.
你怎么来闭环这个循环呢？
How do you close that loop?
对我们来说，有一个办法是，可能有一个固定的地方，反馈通常在那里发生。
A thing that has worked for us is there's probably some kind of place where this feedback usually happens.
通常是当面说，还是在别的地方？
Is it usually in person or does it usually happen in...
我们所有人都在用 Telegram，但 Claude 没有原生集成。
For all of us on Telegram, there isn't a native Claude connection.
我们有个机器人会抓取所有聊天记录，但还是得人工操作。
We have a bot that scrapes every chat, but it's still very manual.
我得手动让 Claude 去看机器人和那些反馈。
I have to trigger Claude to look at the bot, look at the feedback.
而且这还不算……
And it's still not...
还没完全到位。
It's it's not there.
就像，连我们发过的东西都看不好。
Even, like, looking at something we've sent.
因为我们确实，显然……
Because we do we do obviously...
再说，在编码的语境下，我们有很多工程师，你大概知道哪个工程师适合评审你正在做的某个东西。
Again, in a coding context, we have a lot of engineers who you sort of know which engineer you want to review a certain thing that you're working on.
而他们会关注什么，Claude 很擅长从反馈中推断出来。
And the kind of things that they look for, Claude is very good at figuring out from the feedback.
它很擅长，比如，根据我读过的所有这些反馈。
It's pretty good at, like, given all of this feedback that I've read.
Slack 上的反馈。
Feedback on Slack.
在 GitHub、Slack，或者其他任何可能有反馈的地方。
On GitHub, Slack, wherever else the feedback might happen.
你并没有明确告诉它要看反馈，只是给了它 Slack 的权限，它就会自己找反馈？
And you didn't tell it to look at feedback specifically just because you gave it access to Slack, it's looking for feedback?
因为它总是会……
Because it will always...
这样容易多了。
That is much easier to do.
比如，如果它有这些工具，你就不用自己费劲去整理所有给过的反馈。
Like, if it has the tools and you don't have to, like, as a human yourself, figure out, okay, let me assemble a list of all the feedback I've ever given.
但如果你按三步走，第一步，整理所有你给过的反馈。
But if you do this like three step process of step one, assemble all the feedback you've ever given.
第二步，让Claude把这些反馈浓缩成你在意的规则或原则。
Step two, tell Claude to like condense that down into a list of rules or principles that you care about.
有时候，再来第三轮也挺有用的，比如，看看有没有一些你可能会相信但自己其实没意识到的原则。
Sometimes it's useful to like then again do a third round of, like, okay, what are the principles that maybe you might believe but you don't actually know you believe?
然后把这个清单作为你后续评审的参考。
And then take that list as, like, an input for your ongoing reviews.
这个方法在我们尝试让每次会话的反馈更专业方面，效果相当不错。
That has that has been quite successful in our attempts to, I wanna say, like, specialize the feedback that individual sessions give.
真的挺管用的。
That has really worked.
我觉得对于这种工作，跟纯写代码或者不太依赖文字的工作相比，问题可能更大一些，但我觉得类似的方法还是可以用的。
I think for for work that is a little different than raw coding or not very text based, I think the problem is still bigger, but I think similar things can be done.
比如，给我反馈一下客户通话的情况。
About, like, give me feedback on a customer call.
那通电话打得怎么样？
How did that go?
如果你读了通话记录，或者比如你可能……
If you read the transcript or, like, maybe you're...
我最近在帮Anthropic做很多面试，这算是典型的知识性工作问题，但也挺有意思的。
I'm doing a lot of interviews for Anthropic, which is, a typical knowledge book problem but also, like, quite interesting.
但那样的话是它在教我。
But then it's teaching me.
我想让它自己学。
I want it to teach itself.
我觉得很多刚开始用AI的人都有这种体验。
And I think a lot of people who started using AI experience this.
你会想做更多。
You want to do more.
你会看到自己有多高效。
You see how productive you are.
但你……
But you...
我不是……
I don't...
没有一个智能体是百分之百让我信任的。
There is no agent that I trust 100%.
每一份输出都要经过我们团队，最后还要到我这里。
Every output goes through my team and then eventually through me.
所以我成了公司里最大的瓶颈。
So I become the biggest bottleneck in the company.
说实话，我觉得我们的产品现在就能帮你很好地解决这个问题。
Honestly, I feel like our products would do that for you pretty well today.
我其实挺有信心的。
I'm actually quite confident.
比如说，如果Claude能看着你给反馈一段时间，他可能首先会做的就是把两者放在一起对比一下。
Like, if Claude has the ability to watch you give this feedback for a while, like, the first thing he would probably do is he would sort of run it side by side.
这个实验我几乎推荐给所有人。
This is an experiment I recommend to almost everyone.
做起来真的很有意思。
It's like really fun to do.
如果你现在有信任问题，你可能不会特别兴奋地立刻交出控制权，对吧，就像从一天到另一天那样。
If you currently have a trust problem, you will probably not be super excited to immediately give over the reins, right, like from one day to the next.
但可以试着并行运行一下。
But sort of run it side by side.
比如，读点什么，写下你的反馈，这样来几轮，就是为了看看Claude会给出什么样的反馈。
Like, read something, you write down your feedback, and you do that for a few rounds, and that's just to check what kind of feedback Claude would have given.
然后看看，它们哪里有不同，哪里又是一样的？
And like, where do they differ, where are they the same?
这也让你有机会稍微优化一下工作流程，因为我觉得模型还是会有不完美的地方。
That also gives you something where you can actually sharpen the workflow a little bit, because I think models will still be imperfect.
还是会有一些事情……
There's still gonna be things that...
就是反馈会有点不一样的那些地方。
Where where the feedback is, like, a little different.
而且，即使对大多数人来说，如果让同一个人在不同的时间，比如周一和周五，去评审同一份文档，得到的反馈也会不同。
And, like, even for most humans, if we take the same human and we ask them to review the same document, like, on a Monday versus a Friday, the feedback will be different.
但这种并排对比的过程其实挺强大的，因为它能让你看到很多东西。
But this, like, side by side process is is rather powerful because it shows you so much.
那我怎么构建这个能和团队对话的代理呢？
And how do I build this agent that talks to my team?
它是那种……
Is it just something that...
比如，一个网站，团队成员上传内容，然后它给出反馈？
Like, a website where they upload stuff and it gives feedback?
我推荐用Claude Tag，也就是Slack里的Claude。
I would recommend Claude Tag, which is Claude inside Slack.
那你建的是...
And you build...
你让它了解你的品味。
You give it access to my taste in it.
这就像我们目前核心的多人在线体验。
It's like it's like currently our core multiplayer experience.
也就是多个人可以同时和Claude对话。
Where multiple people can talk to Claude at the same time.
在构建东西这方面，这一切变得非常有趣。
This is all getting very interesting when it comes to building things.
因为你可以很容易地构建很多东西。
Because you can build so many things easy.
当你开发软件的时候，什么变得有价值？
What becomes valuable when you're building software?
我说的是那些刚开始创业的人。
And I'm talking about people who are beginner entrepreneurs.
我在想我自己。
I'm thinking about myself.
那是什么呢？
What is it?
2015年，我们带着一个留学体验的预订平台来到这里。
2015, we came here with a booking platform for study abroad experiences.
太棒了。
Amazing.
当时，我们因为雇佣了很多工程师，投入了几十万美元来搭建这个平台，自动化了很多东西，形成了竞争壁垒。
A moat was building this whole platform because we hired a lot of engineers, was down hundreds of thousands of dollars, we automated stuff.
现在我自己两天就能搞定这些。
Now I can do this in two days by myself.
那么，在考虑开发软件的时候，我的创业方式是什么呢？
Like, what is my entrepreneurial mode when you think about building software?
有吗？
Is there any?
这是我的观点。
Spiese opinion.
我觉得其实变化没有那么大。
I think it actually hasn't changed that much.
我认为投资成本，比如前期资本投入，以前要高得多。
I think the cost of investment, like sort of the upfront capital cost, was much higher.
确实是这样。
That is true.
而且，我觉得我们现在正进入一个有趣的阶段，一个人或两个人的团队就能打造出价值十亿美元的公司。
And, like, I think we're now entering this interesting area where, like, single person teams or, like, two person teams can build billion dollar companies.
但我觉得打造好产品的秘诀其实没有变。
But I think the secret to building a good product has actually not changed.
我相信那就是同理心。
And I believe it's empathy.
对用户有同理心，对使用场景有同理心，还要去思考这个东西对最终愿意付钱的人到底有没有价值。
Empathy for the user, empathy for the use case, empathy for like trying to figure out is this actually valuable for the person who's ultimately going to pay for it.
我觉得这一点……
I think that has...
再说一遍，这也许是个比较尖锐的观点，但我认为这向来是创业团队的关键要素。
Again, and I think this is maybe a spicy opinion, but I think this has always been the key ingredient in any entrepreneurial team.
而且我觉得在创业公司里尤其如此，比如在我们开始录制之前，你和我聊了一点关于500 Startups的事情。
And I think especially in startups, like before we started taping, you and I talked a little bit about 500 startups.
我觉得这在风投圈里几乎是常识了，你拿什么出来其实并不重要。
I think this is even like sort of common knowledge within VCs, that whatever you show up with doesn't really matter.
反正它大概率会变的。
It's probably gonna change anyway.
但作为创始人，你才是关键。
But you as the founder, you're the key ingredient.
我们还可以再往深一层说。
And we can we can go one level deeper.
那是什么意思？
What does that mean?
是说你的才华、你的技能吗？
Is it like your talent, your skills?
我觉得真正能做出有价值的东西，加上同理心，可能才是创始人最核心的特质。
I think empathy and like an ability to really build something of value is probably the core thing that matters in a founder.
而且这一点一直没有变过。
And I think that has not changed.
只是现在要发挥这种同理心，去执行起来容易多了。
It has just gotten much easier to tap into that empathy and, like, to to execute on it.
我完全同意。
I totally agree with that.
但有时候我也会想，我有个想法，结果在X上看到有人说，哦，这个新模型直接把这个淘汰了。
But also sometimes, I'm thinking, like, I have this idea, but then I see something on X where people say, oh, this new model just killed this.
而且我们看到大家用 vibe coding 做小工具之类的。
And we see people vibe coding their smaller tools, etcetera.
你觉得我们会越来越倾向于自己随便搞个工具，而不是花钱去买现成的工具吗？
Do you think it's a trend that we're gonna vibe code a lot of tools versus paying for them?
比如，如果花一个小时就能做出类似的东西，那还有什么理由让我们愿意为某个工具付费呢？
Like, what would make us pay for a certain tool if we can spend an hour just building something similar?
我觉得那些注定会消失的软件，都是糟糕的软件。
I think software that is hopefully gonna go away is bad software.
就是那些根本不适合你需求的软件，你之所以用它，只是因为没有什么更好的选择。
Like, the software that doesn't really serve your use case and you're only using it because you don't have anything better.
我觉得每个人都能想到一些工具，你用的时候其实不太情愿，心里想着：算了，还是用它吧，因为不得不用，也没时间去找到更好的替代品。
I think all of us have some tools that come to mind where you're like, begrudgingly, okay, I guess I'm gonna use this because I have to and I don't have the time to like look for something else or use something else.
这种状况会结束的。
I think that will come to an end.
不过同时，我们也提到过，做出好产品有多难。
However, at the same time, we talked about how hard it is to to build good products.
我觉得找到最佳的方式来传递价值、带着同理心去思考，依然会有巨大的价值。因为大部分我真心喜欢的那些产品，我自己是想不出来的。
I think there's still going to be tremendous value in, like, figuring out how to best deliver value and how to, like, to have empathy in a way where most of the products I use that I'm, like, a big fan of, I wouldn't have come up with myself.
我说过，我写过博客文章说iPhone很蠢，因为显然你需要一个键盘。
Like, I'm on record as, like, having written blog posts about how the iPhone is stupid because obviously you need a keyboard.
哦，
Oh,
真的吗？
really?
当然。
Of course.
我觉得我真正喜欢的大多数产品，都是别人想出来的主意，然后做出来，最后给了我。
I think for most of the products that I really enjoy, someone else came up with an idea, executed on it, and then gave it to me.
我觉得我使用这些产品的主要理由，很少是因为它们做起来很贵，或者花了很长时间，或者需要一个大团队。
And I think the main reason I use those products is rarely because it was very expensive to build or because it took them a long time or because they needed a big team.
我使用某个产品，是因为它能很好地解决我的需求，可能比我自己想出来的方式更好。
The reason I use the product is because it solves a need I have really, really well, probably in a way that was better than I thought of.
我觉得我们将会完全填补的空白，就是那些垃圾软件，我不得不用它们，因为我没有其他选择。
The gap that I think we're gonna completely fill is this garbage software and I only have to use it because I have no other choice.
我觉得这 hopefully 会消失，我对那个未来非常期待。
I think that will hopefully go away, and I'm very excited about that future.
在我们继续之前，我想多了解你一点。
Before we keep going, I'd love to get to know you a little better.
我在这里分享了很多我的工作，我也很好奇你的。
I share a lot about my work here, and I'm curious about yours.
你是做什么的，你希望在哪方面得到更多帮助？
What do you do and what you'd like more help with?
我们准备了一个小调查。
We've put together a little survey.
你的回答会帮助我决定未来对话要涵盖什么内容，以及为你准备哪些资源。
Your answers will help me decide what to cover in future conversations and what resources to put together for you.
我会把链接放在描述里。
I'll leave the link in the description.
拜托拜托拜托，花几分钟时间告诉我关于你、你的生意、你的抱负，这样我就能为你制作更好的内容。
Please, please, please take a couple of minutes and tell me about you, about your business, about your ambition so I can make better content for you.
我真的很感谢你。
I really, really appreciate you.
非常感谢你抽出时间来做这件事。
Thank you so much for doing this.
你提到了一些关于静默运行工作的Windows电脑，以及它们是一个巨大的商业机会。
You mentioned something about Windows computers that quietly run work and that they are a big business opportunity.
你能跟我聊聊这个话题吗？
Can you talk to me about that?
我做AI产品。
I build AI products.
我做Claude。
I work on Claude.
当我走进现实世界，观察真正支撑经济的电脑和机器时，我觉得我们硅谷的人很容易只想到那种用着一堆最新款MacBook的典型创业公司，好像世界就只是这样。
When I go out into the world and look at what are the computers and the machines that are really running the economy, I think it's really easy, especially for us in Silicon Valley, to consider your typical startup with a bunch of latest generation MacBooks, and that's all that exists.
然后你去机场，往角落里一瞥，会发现他们其实还没用上图形操作系统呢。
And then you go to an airport and you peek around the corner, and you see that they have not actually yet moved to graphical operating systems.
还是那种黑白的界面。
It's still like the black and white like
哦，对。
Oh, yeah.
交通管理部门。
The transportation authorities.
现在有时候我觉得好玩，就专门去那些地方，那些……
And I'm doing this for fun now sometimes that I just go to all of these places that I...
那些支撑着经济运转的地方。
That that run so much of the economy.
比如我上次去的就是一家酒店。
Like, the last place I did this was a hotel.
我喜欢在办理入住的时候这样问：你好，我做软件的，开发工具。
Like, I love doing this at a check-in where I'm like, hi, I work in software, I build tools.
能不能让我看看你手边那台小机器上是怎么给我办理入住的？
Can you just show me real quick what you do on your little machine right there to check me in?
太吓人了。
Harrowing.
所有设备里的软件都特别笨重，特别糟糕，真的差劲，我觉得这对很多企业来说可能是一个未被开发的机会，进去把这些东西清理干净。
All devices of have software that is so clunky, so bad, truly awful, and I think it is probably an untapped opportunity for many businesses to go in there and, like, try to clean that up.
早在2010年AI兴起之前，我就开玩笑说，如果我想退休，我就给牙科诊所开发真正好用的软件。
Well before AI became a thing in 2010, I I joked that if I ever wanted to retire, I would build really good software for dentist offices.
现在你大概一天就能做出来了吧。
Now you can build it in a day, I guess.
难的是人际关系那部分。
It's the relationship part.
对，没错。
Yeah, exactly.
那正在变成……
That's becoming...
嗯，那是最难的部分，对吧？
Well, that is the hardest, right?
不只是软件的问题。
Not just the software.
你还希望有什么样的企业存在？
Any other businesses you wish existed?
我仍然觉得我们在让软件更具个性化、带有一点AI味道这件事上才刚刚起步，而不只是加一个带聊天机器人的侧边栏。
I still think we're barely getting started on software that feels a little bit more personal and has a little bit of a touch of AI that is not just here's a here's a sidebar with a chatbot in it.
我觉得我们基本上用那些低智能、便宜的模型把整个行业都渗透了，这些模型基本就是读了FAQ页面，你一问问题，它就能把答案吐出来。
I think we've we've sort of permeated the entire industry with low intelligence, cheap models that operate sort of at the level of I've read the FAQ page and if you give me a question, I'm sort of like, I can spit out the answer.
我们还没有真正弄清楚，如何在很多应用中充分利用顶级智能。
We haven't really like figured out what can we do with top line intelligence across many of our applications.
至于我特别希望存在的企业，我就再次强调一下同理心吧。
Businesses that I specifically wish existed, I'm just gonna double down on this empathy thing.
这个行业变化很慢。
The industry moves slowly.
连风投都慢吞吞的，适应新世界真的很难。
Like even VCs move kinda slowly and it's like really hard to adjust to a new world.
我觉得我们还没有真正适应执行成本变得廉价这件事。
And I don't think we've properly adjusted to execution is cheap.
我们一直说想法是廉价的。
We've always said ideas are cheap.
这倒是真的。
That was true.
我们也说过执行成本有点高，但随时间推移变得越来越便宜。
And we said execution, kind of expensive, but it's it's gotten cheaper over time.
现在执行成本已经变得极其便宜了。
And now execution has become extremely cheap.
我觉得我们需要真正加倍重视同理心和连接，看看谁能深入业务，真正理解需求在哪里，以及怎么...
I think we need to like really double down on empathy, connections, who's capable of like going into a business and really understanding what the needs are and how to can...
我们在这里到底怎么优化事情。
How can we actually optimize things here.
所以这可能真的是个好机会。
So this might actually be a nice opportunity.
更大的复仇其实是那些MBA们。
Like, the bigger revenge are the MBAs.
我觉得MBA们现在可以创办很多软件初创公司，而且可能会相当成功。
Like, I think MBAs can now make a bunch of software startups that might be quite successful.
我刚和一位诺贝尔经济学奖得主聊天，他告诉我，我们有如此强大的模型，但真正缺少的是应用层。
I was just talking to a Nobel Prize laureate in economics, he was telling me that we have such powerful models, but we really lack the application layer.
你同意他的看法吗？模型的能力和普通用户的使用之间，存在这么大的差距吗？
Do you agree with him that there is this huge gap between what models are capable of and what average user is using them for?
哦，是的。
Oh, yeah.
当然。
For sure.
我们确实存在能力过剩的情况。
We definitely have, like, a capability overhang.
比如，你面前有一个很棒的模型，而你基本上只是用它来问，嘿，法国的首都是哪儿来着？
Like, the idea of you have an amazing model in front of you and if you're essentially using it to to ask, hey, what's the capital of France again?
你很可能没有真正挖掘出它的潜力。
You're probably not really tapping into what is possible there.
我们的创始人和CEO，达里奥，提到了一个概念，就像是“天才数据中心”一样。
Our founder and CEO, Dario, talks about this concept of, like, a data center of geniuses.
我的工作实际上就是想办法，好吧，假设我们有一个天才数据中心。
My job really is to figure out, okay, like, let's assume we have a data center of geniuses.
我该如何让你，作为用户，真正利用好它呢？
How do I empower you, the user, to actually make use of that?
这需要创造力。
And it requires creativity.
这需要重新思考你的需求和你认为可能的事情。
It requires, like, a rethinking of, like, your needs and, like, what you think is possible.
这是一个非常棘手的问题，但我认为它并不新鲜。
It's a it's a really hard problem that I don't actually think is new.
人们之所以受限，并不是因为实际可能性，而是因为他们所认为的可能性。
It's this notion of people have always been held back by they thought is possible rather than what is actually possible.
我确实觉得我们只是刚刚触及了可能的表面。
I do agree that we have, like, we've just scratched the surface of what's possible.
个人层面上，怎么能缩小这个差距呢？
How can one close that on an individual level?
个人层面？
On an individual level?
这个问题我整个周末都在想。
I thought about this, like, all weekend.
我觉得在个人层面上，你经常能看到一种号召，就是要求人们要有更大的野心。
I think on an individual level, the thing that you see on acts a lot is this this rallying cry of be more ambitious.
我们应该更有雄心。
We ought to be more ambitious.
但更有野心到底意味着什么，我觉得非常取决于具体情况。
Like what being more ambitious really means is I think so context dependent.
但我希望大家能摆脱所有对可能性的束缚，重新思考自己为什么要做这些事情。
But I would like people to shed all the shackles of what they believe is possible, and rethink why are they doing things.
比如说……
Like what...
比如我在业务里做某件事。
Like I'm doing a certain task in my business.
我希望接下来发生什么？
What I what am I hoping for is going to happen next?
然后你就一直给模型这个任务。
And you just keep giving the model that task.
不管你想要什么结果，就把那个任务交给模型。
Like whatever outcome you're hoping for, give the model that task.
你不能一直这么玩下去。
You can't play this forever.
你最终会遇到当前最强智能的极限。
You will eventually find the limits of the current top intelligence.
一个有力的反驳可能是，大多数企业的目的是赚钱，而如果我只是跟Claude说，去赚钱，它可能走不了多远。
Like a good counterargument here is, okay, like, the purpose for most businesses is to make money, and if I just tell Claude, go make money, it's probably going to not make it very far.
所以我们还没达到最高的高度。
So we we haven't reached, like, the top level of altitude yet.
还在中间某个位置。
It's somewhere in the middle.
但一旦你碰触到那个天花板，你就会知道自己已经站在最前沿了，你就是那个先锋，而且你会从下一代模型中真正受益。
But once you bump against that ceiling, you do know that you're at the forefront, you're the frontier, and You're you're gonna really gonna benefit from the next model generations.
因为如果我们做的只是帮我总结这封邮件，或者让这封邮件听起来不那么生气，那你并没有真正从新的人工智能中获益。
Because if all we're doing is summarize this email for me or make this email sound less angry, you're not really benefiting from the new intelligence.
对于现在刚开始从事软件工程的人，面对这个不断变化的领域，你有什么建议？
What would be your advice to someone who's just starting out now as a software engineer in this ever evolving?
在这个不断变化的软件开发领域中。
In this ever evolving landscape of how to do software engineering.
我觉得这真的取决于你作为软件工程师想要什么。
Think it really depends on what you want as a software engineer.
比如，做一名软件工程师意味着什么？
Like, what does it mean to be a software engineer?
我觉得我们作为工程师，现在正处于一个危机的时刻，我们在想，做一名软件工程师到底意味着什么？
I think we're, like, as engineers, currently in this moment of crisis where we're like, does it mean to be a software engineer?
是不是意味着我……
Does it mean that I'm...
我已经把 React 的整个 API 服务都背下来了吗？
I've completely memorized the entire API service in React?
这就是我的价值所在吗？
Is that what I'm valuable for?
这就是我干的活吗？
Is that what I do?
我觉得现在的情况可能已经不太一样了。
And I think that's probably less the case now.
对我来说，这是一个挺受欢迎的转变。
To me, this is a pretty welcome change.
我一直认为，一个好的工程师和一个差的工程师，我衡量……
I've always been of the opinion that a good engineer versus a bad engineer, I measure...
我一直以影响力来评价工程师，所谓影响力就是，用户真的喜欢到愿意花钱买这个软件吗？
I've always measured my engineers on their impact, and the impact is like, does the user like this enough to buy the software?
对我来说，那些特别优雅、精美且技术高超的方案，一直以来价值都不大，即使我们对工程的工艺感到自豪，但用户没感觉到任何好处。
There's there's always for me been relatively little value in, like, really elegant, beautiful, mastery solutions where we can be really proud of the engineering craft here, but the user doesn't feel any benefit.
你是怎么培养这种能力的？
How do you develop that?
现在大家都变成了资深职位，那么为了在三年后还能保持竞争力，你会在哪些方面加倍投入？
What do you double down on now in order to be relevant in three years if, like, every position is suddenly a senior position?
你在这方面有什么建议吗？
What would be your advice there?
我当初进入这一行，靠的是建立强大的作品集，我觉得现在这依然是最好的建议。
The way I got my start in this career is by, like, building a strong portfolio, and I think that would be the same advice today.
我认为，建立一个不仅仅是模拟项目的作品集，现在已经容易得多。
I think building a portfolio that is not just like a mock is much easier.
所以我对年轻人的建议就是，去做一些真实的东西。
So my recommendation for anyone young would be go and build some real things.
证明你能做到。
Prove that you can do it.
证明你有同理心。
Prove that you have the empathy.
证明你有真本事……
Prove that you have the chops...
能够跟客户沟通。
The ability to talk to customers.
搞清楚什么必要，什么不必要。
Like, figure out what is necessary versus not necessary.
我确实觉得，软件工程的定义有点变了。
I do think there's, like, a slightly new definition of what software engineering is.
我个人觉得这是好事。
I personally think that's good.
我确实看到，年轻人适应这种新思维比一些老工程师快得多，那些老工程师可能还享受那种主要就是做任务、不需要问太多为什么的安逸。
I do see that young people adapt to this new style of thinking much quicker than some of the older engineers who maybe enjoy the luxury of primarily working on on tickets without having to ask why too many times.
我见过很多人在那方面做得非常成功。
I've seen so many people be quite successful in that.
相比之下，我现在可能不太推荐的那些事情。
The things that I would maybe, to contrast, not recommend anymore.
确实，我觉得完全把每一个API服务都背下来，价值已经没那么大了。
It is true that I think, like, completely memorizing individual API services, I think, has become less valuable.
真正有价值的，还是能把不同的系统以有趣的方式结合起来。
What is still valuable is being able to combine different systems in interesting ways.
我觉得现在这种能力被放大了，因为你可以很快地把东西组合起来。
I think the value of that is amplified now because you can so quickly combine things.
你只要指一下AI和三个不同的系统，然后说，我想把这几个……
You can just, like, point AI and, like, three different systems and be like, I want these three...
把这些五个东西混在一起，去做吧。
Like, want these five ingredients mashed together, go and do it.
那你在招人的时候会关注这个吗？
Is that something that you look for when you hire people?
这种能力……
This ability to...
你还有其他特别看重的东西吗？
Is there something else that you're particularly looking for?
我不知道你是怎么招聘的。
I don't know how you hire.
你还……
Do you still...
你还是很看重简历吗？
Do you do you care a lot about the resume?
对。
I do.
说到……
When when it comes...
因为你提到人际关系，这在我的行业里非常重要。
Because when you mentioned relationships, they're very important for my industry.
我在找那种人脉广的人。
I am looking for someone who knows people.
我在找那种了解其他公司怎么运作的人，因为这些信息不是公开的。
I am looking for someone who knows how other businesses operate because this is not public information.
比如，我怎么才能知道...
Like, how do I learn how to...
我作为一个播客主，应该收多少钱？
What what do I charge as a podcaster?
什么？
What do
我还收蚂蚁的钱啊？
I charge for ants?
我想要一个对市场情况了如指掌的人。
I want someone who knows exactly what's going on in the market.
所以我们确实很看重人脉。
So we do care about relationships.
但说到创意部分……
But when it comes to the creative part...
其实，创意部分我们还是看他们的视频有多少浏览量，因为确实是这样。
Actually, with creative part, we still look at how many views their videos have produced because Exactly.
他们懂艺术。
They know the art.
当然，我不看他们的学历。
I don't look at their degree, of course.
没错。
Exactly.
没错。
Exactly.
我觉得这是……
And I think this is...
我不知道这是不是那么有争议。
I don't know if this is so controversial.
我想我们俩的工作都算是科技相关的，但我从来不太在意你有什么学位或上过什么学校。
I guess we both work like in a, you know, tech adjacent, but I've never cared too much about which degree do you have or which school you went to.
对我来说，更容易的是看你自己做过什么东西。
Was always much easier for me to look at what are the things you made.
那些东西我可以亲手操作、感受、尝试，然后我就会觉得，啊，你要么是队里最能赢的那个人，要么至少在一个能赢的团队里干过。
Those things I can touch, feel, try, and I can feel like, oh, you either were the winning player on a team or you at least worked on a winning team.
这有时候对我来说挺有用的。
That is sometimes useful for me to know.
我觉得那真的是很有价值的信息。
I think that's really valuable information.
而且问题是，如果没有这些，我可能会走捷径去看学校、看学位。
And the thing is, if I didn't have that, I might go take the shortcut and look for schools, degrees.
我其实不用那么做。
I don't really have to.
我总会选一个已经证明自己能做出好东西的人，而不是那个没做出什么好东西但拿了个好学位的人。
I will always pick someone who's proven themselves to build something amazing over someone who hasn't built something amazing but has an amazing degree.
你会怎么跟你女儿说，如果她长大后对你说：爸爸，我想学编程。
What would you tell your kid when she grows up and says like, dad, I wanted a coding.
你会给她什么建议？
What would be your advice?
她现在才九个月大，所以我觉得……
I mean, she's nine months old right now, so I think
我觉得事情会变得不一样。你女儿可能已经要排队上幼儿园了吧。
I things will change a little think you have to be in line for your daycare already.
我就是说说而已。
Just saying.
而且一切发生得比你想象的要快。
And it all happens quicker than you think.
我会跟我女儿说什么呢？
What would I tell my daughter?
我想我会告诉她，要尽早去理解企业是如何运作的，人们是如何协作的，以及如何真正找到价值。
I think I would tell my daughter to very early on try to understand how businesses operate, how people collaborate, how to really find value.
我没去过滑铁卢大学，但我非常欣赏他们的一套做法。
I have not gone to the University of Waterloo, but I'm such a fan of the University of Waterloo's process.
那么多优秀的人才。
All the amazing minds.
确实，那么多优秀的人才。
All the amazing minds.
不过我觉得滑铁卢大学一直以来在计算机科学和工程领域有个很特别的做法，这算是我的一个外部视角。
But I think the University of Waterloo has always had this approach which was quite unique within within computer science engineering, which was this is an outside perspective.
他们可能不这么认为。
They might disagree with me on this.
我的外部视角来自于德国，在那里计算机科学就是纯粹的计算机科学。
But my outside perspective was always coming from Germany where computer science is very much computer science.
是信息学。
It's informatics.
很多数学，而关于怎么真正做出好产品，就只有一丁点，更别提计算机工程了。
It's a lot of math and very little, like only a sliver of how do we actually build a good product, let alone computer engineering.
在德国，JavaScript 长期以来都不被看作是真正的计算机科学家会用的东西。
JavaScript was in Germany for a long time universally considered to be like not what real computer scientists use.
滑铁卢大学的方法一直让我很感兴趣，因为我遇到的那些实习生或初级工程师，他们之前在湾区已经做过五份实习了。
And the University of Waterloo's approach has always fascinated me so much because I would get these interns or junior engineers who worked at five different companies before in the Bay internships.
在旧金山，尤其常见的是你招一个实习生，这已经是他第四次实习了，他之前在特斯拉、谷歌和苹果都实习过。
It was pretty common, especially in San Francisco, that you get an intern, and it's their, like, fourth internship, and they previously interned at Tesla, Google, and Apple.
他们学到了很多关于公司怎么运作的东西，有哪些优势和劣势，怎么去争取一个想法，怎么获取资源，怎么说服法务部门把事情办成。
And they have so many interesting things they learned about how companies operate, what are the ups ides and downsides, how do you actually advocate for an idea, how do you get resources, how do you convince legal to get something done.
我觉得正是这些能力，让滑铁卢大学的工程师几乎一上来就能发挥很大作用。
I think those are the skills that really for me made University of Waterloo's engineers almost immediately effective.
他们可能是我这里的第一份工作，但他们和那些尽管有计算机科学博士学位但从未在公司工作过的人相比，完全不是一个层次。
It's their first job maybe with me, but they're in a completely different category than someone who has maybe even gotten a PhD in computer science but has never worked in a company.
他们可能在自己的领域很聪明，但就像我们都知道的，打造一个好产品并把它推出去，不仅仅是知道怎么写代码那么简单。
They might be very smart at what they do, but as we all know, like building a good product and getting it out there is about much more than just knowing what code to write.
你需要会和人打交道。
You need to be able to talk to people.
你得学会和其他部门协作。
You need to be able to work together with other functions.
你得对公司怎么运作有个大概了解。
You need to have a little bit of a sense for how does this company run.
所以我会告诉我女儿，真的要在这方面下功夫。
So I would tell my daughter to probably really invest in that.
说得好。
Love it.
那你跟你孩子怎么说的？
What do you tell your kids?
在这个不断变化的经济里，你觉得你孩子应该做什么？
What do you think your kids should do in the ever evolving economy?
我告诉他们什么都去试，我跟他们说你可以成为任何人，因为有时候他们会担心人工智能会抢工作。
I tell them to try all the things, and I tell them you can be anything because sometimes they're like, oh, robots.
你可以成为...的女王。
You can be the queen of...
比如我跟她说，你可以造个机器人，她其实不太兴奋。
Like, if I tell her, like, you can build a robot, she's, like, not very excited.
我就告诉她，你可以当机器人女王。
I tell her, you can be the queen of robots.
然后她就说，哦，我喜欢机器人女王。
And she's like, oh, I like queen of robots.
然后她说，我们买了一辆可以自己编程的自动驾驶汽车。
And then she says, and we we bought this self driving car that you can program yourself.
哦，那挺有意思的。
Oh, that is fun.
她和她爸爸一起组装那辆车。
And she was building it with her dad.
有意思。
Fun.
就像告诉她们，她们什么都能成为，没有什么事情是她们不能碰的，因为那要么是超出认知的，要么就是别的什么。
And, like, just telling them they can be anything, and there is no thing that they cannot touch because it's, I don't know, out of this or whatever.
因为我初来乍到时，觉得这里的每个人都是外星来的。
Because I came here with a notion that everyone here is from another planet.
如果你上了斯坦福，对我来说，那简直是我想都不敢想的。
If you went to Stanford, for me, that was, like, something that I couldn't even fathom.
但看到这些都是普通人，只是非常热情、勤奋，而且把这个想法传递给我的女儿们，让她们相信一切皆有可能，这只是努力和热情的问题。
But seeing that these are all normal people, just very passionate, hardworking, and, like, giving this idea to my daughters that anything is possible is just hard work and passion.
哎呀，天哪。
Oh, man.
我喜欢这个。
I love that.
这个...
The the...
我之所以这么喜欢，是因为在人工智能圈里，我们经常讨论野心这个概念，有些人野心不够，是因为我们生来就被设定成只会考虑某些可能性，如果觉得不可能，我们就连试都不试。
I think the reason I love that so much is because in AI circles, we constantly talk about this idea of ambition, and the reason people are not ambitious enough is because we're sort of programmed as humans to, like, only consider certain things possible, and if we don't consider them possible, we don't even try.
所以这真的很...
So it's like real...
就是说，我们相信什么是可能的，才是真正的限制。
Like, what we believe is possible is the real limiter.
比模型能做什么或者我们能做什么，限制要大得多。
Like, much bigger limiter than what the models do or what we can do.
也许这也是我特别相信新一代的原因，因为他们从来没有经历过我们这些限制。
Maybe that's another reason I really believe in the new generation is because they will never have experienced all these limitations that we have.
好的。
Okay.
我觉得我们聊得真的很深入了。
I think we covered so much.
我就想听听你对未来的最后想法。
I just wanted your final thought on our future.
你觉得五年后我们的生活会有哪些不同？
How do you think our life is gonna be different in five years?
这个问题也太大了。
That is such a big question.
我相信你找对人了，比如……
I'm sure you have the right person for, like...
我知道，上任何播客都有一个职业风险，就是非得做预测不可。
I I know that, like, going on any podcast, it's like there's a there's an occupational hazard of having to make predictions.
到目前为止，我注意到最大的一个变化就是，我们对AI带来的智能指数级增长都大错特错了。
Biggest thing that I've noticed so far is that we're all so wrong about the intelligence exponential coming out of AI.
真的很难具体理解，指数级的增长到底意味着什么。
It's really hard to, like, wrap your finger around what does it really mean for for the exponential to exist.
我一直回到数学上来想。
And I keep going back to math.
比如，当我们在马拉松竞赛中获得银牌的时候——我说“我们”是指谷歌这个行业——我已经很了解规模法则了。
Like, when when we won the silver medal at the Marathon Theater, when I say we, I mean Google as an industry, I was well aware of scaling laws.
我很清楚智能进步的速度有多快。
I was well aware of, like, how much intelligence is progressing.
而且我想，如果你当时问我，AI什么时候能解决我们认为无法解决的数学难题，我绝不会说2026年年中就会实现。
And I think if you'd asked me at what point will AI solve what we would otherwise consider unsolvable math problems, I don't think I would have given you, like, the middle of twenty twenty six.
我可能会说，那得是一年半以后的事了。
I would have been like, that's that's in a year and a half.
你在说什么呢？
What are you talking about?
所以我觉得，凭直觉去判断未来会怎样……
So I think instinctively having a good feeling for how well...
真的非常难。
What will the future look like is really really tricky.
这是第一点。
That's the first part.
而第二点对我来说也挺难的是，我能比较有把握告诉你的一件事就是，我们现在用的产品，以后都会变得完全不一样。
And I think the second part that is rather tricky for me is the one thing I can tell you with a decent amount of confidence is that I think the products that we use today will all look completely different.
我觉得我们现在正处在……
I think I think we're currently...
我们已经到了一个人工智能变得非常主流、非常受欢迎的阶段，以至于很容易让人以为，哦，可能这就是最终形态了。
We've hit this point where AI was becoming so mainstream and it's becoming so popular that it's, like, easy to imagine, oh, we're maybe like this this is sort of the end state.
但我非常确信，我们今天用的所有AI产品，都跟智能手机出现之前那些奇奇怪怪的手机差不多处于同一阶段。
But I'm quite convinced that all the AI products we use today are roughly in the same phase as all the funny phones that we had right before the smartphone came along.
所以我觉得我们还处在iPhone时代之前，这让我觉得在这个领域工作非常令人兴奋。
So I think we're still in the pre iPhone era, which makes it very exciting for me to work in.
哇。
Wow.
这真的太吸引人了，因为从一开始……
This is really fascinating because already, like, from...
在我们还没进入AI时代的时候，就有那么多东西在变化，你让我重新思考了我整个流程。
In our pre AI era, there are so many things that are changing, and you made me rethink my whole process.
我现在都不好意思给你看我现在的做法，因为我……
I won't even show you what what it looks like now because I'm
我确定没问题，如果……
sure it's fine if
它看起来过时了，因为我们把所有数据给它，告诉它具体要看什么，但我完全明白，让它接触所有信息会强大得多。
it works outdated because because we give it all the data and tell it exactly what to look at, but I totally understand how giving it access to everything is just so much more powerful.
这真是个棘手的问题。
It's such a tricky thing.
我今天给你的所有建议，大概十周后可能就都得更新一遍了。
Like, all the advice I've given you today, in about ten weeks, we're probably gonna have a new revision on all of it.
这倒是真的。
That is also true.
所以对于新模型，我觉得要一直保持与时俱进，不断尝试新东西，这才是关键。
So being an always on agent when it comes to new models, I think that's the that's the thing in trying new things.
非常感谢你，Felix。
Thank you so much, Felix.
如果你和Felix的对话让你想在工作中试试Claude，接下来可以看看我和Ali Miller的访谈。
If this conversation with Felix made you want to try Claude in your work, Watch my interview with Ali Miller next.
Ali和团队合作，教他们怎么用AI。
Ali works with teams on using AI.
在我们的对话中，她展示了自己的实际工作流程，并且直接构建了一个Claude晨间简报。
And in our conversation, showed me her real workflows and built a Claude morning brief life.
如果你想先知道该怎么做，这是个很实用的起点。
It's a practical place to start if you're wondering what to do first.
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。