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。
好的。
Okay.
我们现在在演播室，邀请到了来自OpenAI的Akshay。
We're here in the studio with Akshay from OpenAI.
欢迎。
Welcome.
谢谢。
Thank you.
还有我们的固定搭档Vibhu，听说你们最近推出了ChatGPT Work。
And with our trusty cohost, Vibhu, we so you recently launched ChatGPT Work.
你负责核心产品工程。
You lead core product engineering.
一路走来，能走到今天真不容易。
You It's been a long journey into into all this.
我很感兴趣的是，你最初是从无代码和低代码平台开始的，比如Walrus和Airtable。
I find it very interesting that you started with no code or low code with Walrus and Airtable.
某种程度上，ChatGPT Work 就像是超级应用中的超级应用——终极的无代码工具。
And to some extent, ChatGPT Work is kind of like the super app of Superapps of, well, here is the ultimate no code.
你只要写个提示词就行。
You just write a prompt.
是的。
Yeah.
对。
Yeah.
真有意思，事情居然转了一圈又回到了原点。
It's it's it's funny how things come, like, full circle.
说实话，我整个职业生涯都是这样：一开始做消费金融科技，后来我就在想，如果我们能把工程师用代码实现的功能，以更简单的方式普及给更多人，那才真叫神奇。
I mean, I think for a long time, my career I mean, I started my career working in consumer fintech, but then after that, like, there's this hypothesis that, you know, the things that we were able to do with code, like, as engineers, like, if we could bring that to many more people in a more accessible way, then that would be truly magical.
我们当时在做一家初创公司。
We were working on a startup.
说来有趣，在大模型出现之前，我们就在琢磨怎么用 AI 做自动化测试。
It's actually funny, like, before LLMs, before vision LLMs on how to do automated testing with AI.
那时候我也就是凑合着干，但能做点啥就做点啥吧。
And I was just kinda jank back then, but, you know, doing what we can.
后来我去Airtable待了一阵子，还是秉持同样的想法——如果能把数据库或者数据库背后的基础功能交给普通人用，那会非常有用。
And then and then worked at Airtable for a while on, you know, the same thesis that, like, if can bring a database or the primitives behind a database to people, that would be really useful to them.
但一旦大语言模型出现，我们就清楚了，这正是缺失的那一块拼图，是让普通人无需了解底层原理，就能享受到编程魔力的关键技术。
But once, I think, LLMs came onto the scene, it became clear that, like, this was, like, the missing piece, like, the missing technology required to, like, bring the magic of code to everyone without them having to know what's going on underneath the hood.
所以，这次的发布，还有我们之前做的这些事，本质上都是这个想法的体现。
And so, like, I think this launch and, you know, a lot of the stuff that we've been up to is, like, the manifestation of that.
你刚加入的时候情况怎么样？
How was stuff when you joined?
你是在2023年加入OpenAI的吧。
So you joined OpenAI twenty twenty three.
现在可比那时候丰富多了。
Now we've got, you know, so much more stuff.
比如ChatGPT、Codex应用，还有面向工作的ChatGPT。
So ChatGPT, Codex app, ChatGPT for Work.
情况有变化吗？
Have things changed?
其实我觉得更有趣的是，很多事情根本没变。
Actually, I think the more interesting thing is how things haven't changed.
我记得我刚加入的时候，公司大概只有五百人。
Like, I guess, like, one, I I joined I remember when I joined, it was, like, 500 people.
我当时担心的是，想找一个更早期的项目，会不会真的够初创？
One thing I was worried about was, like, I was looking for something, you know, more early stage and, like, was it gonna feel start up enough?
结果一加入，我就觉得，不错。
And I joined, and was like, good.
这感觉比我能想象的还要更像初创公司。
This feels even more start up y than I could ever imagine.
直到现在，这种氛围一点都没变。
And, like, that that really hasn't changed even till now.
我觉得那种自下而上的野心，还有每个人都能随便提想法、动手实现的能力，真的特别棒。
I mean, I think the, like, level of, like, bottoms up ambition and, like, the ability of anyone to, like, you know, do anything or have an idea and and and ship it is is really cool.
但在使命方面，我觉得最吸引我的是，把前沿智能带给每个人。
But on the, like, sort of mission side, I think what was really compelling to me is this mission of, you know, bringing frontier intelligence to everyone.
也就是打造通用人工智能，然后让它普及到大众。
Like, building AGI and then bringing it to everyone.
而且我觉得，当时他们承认，这个愿景不会是一条直线发展的过程。
And I think acknowledging back then that, like, that vision is gonna, you know, not be a linear progression.
比如，我们可能会尝试不同的产品，有些会成功，有些则会失败。
Like, we're probably gonna, like, try different products and and have different things that that succeed and don't.
但愿景一直没变，使命也没变，我们现在开始看到各个部分逐渐拼合在一起，这真的很酷。
But the vision has stayed the same, and the mission has stayed the same, and we're starting to see the pieces fall together, and and that that's really cool.
你之前做过企业级业务。
You worked on enterprise.
很多人其实都没碰过企业版的 ChatGPT，天哪。
What a lot of people never touch ChatGPT for enterprise, god.
你从那段经历中学到了什么，现在带到了你的工作中？
What is something that you learned from there that you're bringing into your work now?
我觉得企业里根本不存在一刀切的解决方案。
I think how there's no, like, one size fits all solution in enterprise.
我还记得ChatGPT企业版刚推出那会儿，大概是ChatGPT发布一年后，我们跟客户聊的时候，大家都特别兴奋，想把AI引入企业。
I remember in the early days of ChatGPT Enterprise, like, when we talked to customers and, like, everyone that was, like, when I think it was a year after ChatGPT was released, and everyone was so excited to bring, you know, AI into their enterprise.
当时冒出来一堆专门负责AI落地的团队，预算还都特别高。
And, like, all there's all these teams stood up that were being stood up as, like, you know, the AI deployment team with, like, these enormous budgets.
如果你问他们，到底最期待什么？
And if you asked anyone, like, what were they excited about?
他们到底想用AI解决什么具体问题？
Like, what were they excited about solving?
一开始，你听到的都是些标准答案，比如我们有海量数据和信息之类的。
Like, first, you'd get, like, you know, kinda like the the baseline answers of, yeah, we have all this context and data and all this stuff.
但如果你再深入问，具体想让AI在工作中帮他们完成哪一件明确的事，答案就五花八门，完全不一样了。
But then if you ask them, like, you know, what was, like, a discrete use case that, like, they want AI to enable in their in their workplace, you get such a different, like, variance, like, explosion of different types of answers.
挺有意思的，你用这些模型和产品的时候，就像有个万能盒子，想问什么都能问，这才是最神奇的地方。
And it's interesting, like, you know, you using all these models and in these products, you you have this box and you can say anything to it, which is the magic.
但反过来看，这也意味着大家不知道该怎么用它。
But it's on the flip side, it also means that, like, you don't know what to do with it.
在企业里，我觉得很大一部分工作其实是真正走到用户身边，搞清楚他们想解决的具体场景，然后教他们怎么用AI来提升自己的效率。
And in enterprise, I think a big part of that is, like, actually meeting the users where they are, like, use case were they trying to solve, and then actually teaching them how they can use AI to, like, gain leverage there.
你觉得这和前置部署工程有实质性的区别吗？
Do you meaningfully differentiate that from forward deployed engineering?
或者说，我觉得这里面还涉及到市场拓展那部分。
Or I I think there's like the the, like, go to market side of it.
对。
Yeah.
然后是产品方面。
And then there's like the product side of it.
想想新的
Think With new
你需要
you need
去了解
to see
更多关于产品方面的情况。
more of the product side.
对。
Yeah.
我觉得，不管我们在现场支持这类工作上做得多好，归根结底，只要用户盯着电脑或手机，我们的产品就得帮他们找到方向，提供支持。
And I think, like, however good we get at FD emotion, like, I think at the end of the day, if we have a user who's, like, looking at their computer or looking at their phone, like, it's our job in the product to, like, be enabling them and showing them where to go.
所以我们对这一点非常期待。
So we're really excited about that.
你觉得过去三年用户采纳过程中有什么变化吗？
Do you think there's been changes, you know, over the past three years of adoption?
确实出现了质的飞跃，比如有了推理模型之类的。
So there have been, you know, step function changes, you have reasoning models and whatnot.
但企业面临的那些老问题——比如黑箱操作、不知道怎么用——还是没变吗？
Is there still the same problems of enterprise has black box, don't know what to do with it, or have things changed?
我们现在看到，这种增长势头非常迅猛。
I mean, we're seeing now that, like, there's this huge uptake.
对吧？
Right?
大家都对它充满热情。
Everyone's are extremely excited about it.
感觉有好几亿人都在用ChatGPT了。
It feels like, you know, many peep like, hundreds of millions of people are using ChatGPT.
他们已经大致明白该怎么和AI互动了。
They understand, like, how generally to work with AI.
但每次只要一推出新功能，
But then, like, every time, like, a new capability gets unlocked.
比如现在说到智能代理，还是有一批早期用户真正懂行，觉得这东西简直无所不能。
So now, like, we're seeing with agents, like, there is probably a contingent of, like, early adopters still who, you know, truly get it, who are like, you know, you can do anything.
只要你把上下文给对了就行。
You just have to make sure the right context is there.
只要连上合适的工具，再加以监督，几乎什么事都能做到。
It's connected with the right tools, and then you're supervising it, but, like, anything is is is possible.
但还有更大十倍、甚至百倍的市场，只是他们还没意识到，还没看清楚。
But then there's, like, this, like, 10 x or a 100 x bigger market, or, like, they don't yet get that or they don't yet see that.
所以我认为，这将是下一步的发展方向。
And so I think that's the next stage here.
所以，回答你的问题，我觉得用户采纳度已经存在，而且增长得很快，但真正的机会要大得多。
So I guess to answer your question, like, I think the adoption is there and and growing fast, but I think the opportunity is, like, far, far bigger than that.
这正是我们想发力的地方，尤其是ChatGPT Work。
That's where we wanna play, especially with ChatGPT Work.
是的。
Yeah.
那咱们直接跳到ChatGPT Work吧。
Well, let's let's skip ahead to ChatGPT Work.
它是一个月前才发布的，当时是怎么做出这个决定的？
Only, like, a month ago or so announced, What was the sort of decision process that led into it?
你知道，之前有个超级应用的整体整合。
You know, there was this overall merging of the Superapp.
我们正式就这么叫它吗？
Is is that what we're officially calling it?
你们还把浏览器给停用了。
You deprecated the browser as well.
能不能简单总结一下你最近几个月做这个项目的过程？
Just, I guess, summarize your last, like, couple of months of working on this thing.
是的。
Yeah.
感觉像过了好几年，其实才几个月而已。
And it feels like forever now, but I guess it's only been a few months.
我觉得最让人意外的推动点，就是我们推出Codex，或者内部引入Codex的时候。
I think maybe the one one, like, impetus that, like, is most alien is when we release Codex or even internally add Codex.
当时真的让我们很惊讶。
Like, it was really surprising to us.
我们最近发布了一些数据，显示非开发人员在OpenAI内部对Codex的使用出现了明显增长。
I think we recently put out some stats on this, that there's this, like, real inflection of, like, adoption among non developers at OpenAI.
在产品开发过程中，我会参加一些用户研究访谈，跟公司内部的人聊聊。
And I, you know, through this product development process, like, would go to, like, these UXR sessions to talk to people internally.
让我印象深刻的是，不管是战略财务、市场团队，还是其他部门，大家都在用Codex解决自己的实际问题。
And the thing that stuck out to me is, like, one, like, you know, you you go talk to, like, strategic finance or marketing or whatever, they're all using Codex for, you know, their use cases.
这部分挺让人欣慰的。
That that per that part's cool.
但真正让我惊讶的是，大家用Codex时那种自豪感。
But the thing that really stuck out to me is how proud people were that they were using Codex.
就像
Like, how like
这感觉就像
It's like
我不该用它，但
I'm not supposed to be using it by the It
就是这样。
it was that.
他们算是最早接触这个新工具的一批人，而且他们觉得自己获得了某种超能力。
It was like that they were, you know, early to this, like, new thing, but it was also this thing of, like, they felt like they had a superpower.
对吧？
Right?
我们当时意识到，Codex 和智能代理的力量，已经建立在一个庞大的用户基础上——这些人早就熟悉并喜爱上了 ChatGPT。
And what we recognized then is that, like, the power of Codex, power of agents, like, we already have this massive distribution base of people who have, you know, come to know and love ChatGPT.
我们该怎么让他们看到这一点？
Like, how do we show that to them?
我们该怎么把这种能力带给他们？
Like, how do we bring it to them?
这其实是个挺难的产品问题，也很棘手。
Which is like a hard product problem, and it's like a tricky tricky thing.
好吧。
Alright.
这件事有好多办法可以做。
There's many ways you can go about it.
所以我们后来称之为整合，做成超级应用，最终把它融入 ChatGPT 工作模式——具体该怎么实现呢？
And so that's what we call the merge and the super app over time, and and and ultimately launch it in ChatGPT Work is how do we do that?
但这一切源于我们最初的一个认识：这种能力不只是给开发者用的，而且比我们想象的还要早地普及开来。
But it came from that initial realization that, like, the the power was not only for developers, like, much much earlier than probably even we thought.
它其实可以延伸到每一个人。
Like, it could be extended to to everyone.
你觉得这些产品有什么不同？
How do you see the products differently?
那这些产品到底是给谁用的？
So, who is it for?
对吧？
Right?
Codex 一开始连命令行工具都搞定了，接着就是应用程序。
So Codex sorted out even CLI, then app.
现在ChatGPT、Codex和ChatGPT Work已经合并了。
Now there's a merge of ChatGPT Codex and ChatGPT Work.
那这是为普通用户、企业还是工作场景打开的机会呢？
So is it the opening for the average user, for enterprise, for work?
你打算怎么定位它？
How how do you position it?
我觉得我们想把它定位成适合那些做些非传统事情的人，姑且这么讲吧。
I think we want to get it to position it for if you're doing quirky related things, for lack of
找个更贴切的词。
a better word.
对吧？
Right?
我觉得
I think
生产力才是我认为真正的核心支柱。
productivity is like a is is actually what, like, the pillar that I that I support.
这其实就是团队的名字。
Like, that's the name of the team.
我们叫它生产力，而不是企业或者工作之类的，是因为还有个人生产力这一说。
And the reason for that, the reason we call it productivity and not, like, you know, enterprise or or, like, work or something like that is because there's also personal productivity.
对吧？
Right?
我觉得ChatGPT Work，我见过有人在个人生活中用它做些事，严格来说不算工作，但这些助手确实特别管用。
And, like, I think ChatGPT Work is I've seen people do things in their personal lives that you wouldn't classify as like work technically, but like these agents are, you know, super capable for.
最近有个例子，有人在我们的Slack上分享，说他有个包裹没收到。
Like one one recent example that someone posted about on our Slack is like someone has like a missed package.
他看到了亚马逊或者快递公司发来的包裹照片，就让ChatGPT Work帮他查一下包裹在哪。
Like, they didn't receive it, and then they got like the picture of it, you know, from Amazon or whoever the courier was, and they like asked ChatGPT Work to like find out where that package is.
那个助手特别执着，直接看了那张图，然后翻了一堆周边的快递信息。
And like the agent, you know, is extremely tenacious and like like took the image, like looked at a bunch of like listings around neighborhood.
最后准确找到了包裹所在的公寓楼。
I figured out exactly the apartment complex in which the package was.
我给了他们一些信息。
I gave them some information.
我觉得，像这类事情，不管是工作还是提升效率相关的，都挺多的。
And so, like, I think there's all these things that, like, you, you know, work y or productivity related things.
我觉得这正是我们希望产品做到的。
I think that's what we want the product to be.
你提到了 Codex。
You asked about Codex.
我们认为 Codex 是个能长久立足的品牌。
I think we think Codex is, you know, a durable brand.
但我们有个原则：不希望用户被困在某个标签或体验里，用不上产品的核心功能。
But we have a principle that, like, the user you know, we don't want a user to get stuck in a tab or an experience where they don't get the power of the product.
所以，你在桌面端 Codex 里能做的所有事情，在 ChatGPT Work 里也能做，反过来也一样。
And so, like, basically, everything that you can do, you know, in the Codex portion of the product on on desktop, can do and ChatGit with your work and vice versa.
不过我们在产品设计上做了一些明确的取舍，比如：如果你在 Git 仓库里，我们该向用户展示多少 Git 状态信息？
But we made some opinionated product decisions on, like, you know, how much of the Git state, if you're in a Git repo, do we wanna expose to the end user?
或者我们想让‘看到智能体思考过程’的体验做到什么程度，比如逐步展示代码差异，让用户能直观看到这些改动。另外在安全方面，我们该怎么考虑沙箱机制，确保在不同状态下都能设置好默认选项？
Or how much do we wanna make the the experience of seeing the agents thinking, like, diff forward so that you're you get exposed to the diffs out of the and then, like, on the safety side, like, how do we wanna think about, like, sandboxing and making sure that we have the right defaults in one one state versus the other?
所以背后确实有一些产品上的取舍，但我们不希望用户需要自己去选择该用哪种体验。
So there's, like, some opinions that go behind that, but we do we do want we don't want the user to need to choose which experience they're in.
这确实是迈向通用人工智能的一个好目标。
That is a good goal for AGI.
对吧？
Right?
就像，大家并不想自己去选想要哪个版本的AGI。
Like, people don't want, like, to hide to choose what version of AGI they want.
他们只希望AGI替他们做决定。
They just want the AGI to decide for them.
我能得到个明确答复吗？这对我来说有点不太清楚。
Can I get an answer or like, it's not super clear to me?
Codex的工作区和ChatGPT Work的工作区是同一个吗？
Is the Codex harness and the ChatGPT Work harness the same?
这仅仅是界面设计上的区别，还是在提示词层面，甚至更深层的差异？
Is it just UI affordances, or are they actually prompt level or even even deeper differences?
所以这个框架是一样的。
So the harness is the same.
这个框架是共用的。
The harness is shared.
在两个产品里，我们都对框架做了优化，让它更适合知识工作，特别是涉及插件、电脑操作或生成内容的时候。
On in both of the products, we made improvements to the harness to make it good for knowledge work, especially as it relates to plug ins or computer use or artifacts.
不管你用哪个模式，都能享受到这些能力。
You get that power regardless of which experience you're in.
在用户体验方面，我们在Codex模式下有明确的设计主张，比如界面该怎么做、该怎么响应，还有我提到的沙盒相关设置。
On the UX side, there's opinionated takes that we have when you're in Codex mode, what the UX should be how the UX should behave, and some stuff around the sandbox like I mentioned.
但底层的框架和功能应该是一致的。
But the underlying harness and capabilities should be the same.
我只是有点好奇。
I think I'm just kinda curious.
也许我们可以找一个查询，在这两种模式下表现会不一样？
Maybe we can is there a query that we can run that would look different in the the two modes?
是的。
Yeah.
我试过让系统在两种模式下都生成一个退休计算器表格之类的任务。
I tried to create, like to ask it to create, like, a retirement calculator spreadsheet or something in in both in both modes.
在Codex模式下，你可能需要在一个代码库中操作，但你会看到它生成的表格、文件修改这些具体的变动细节。
And then in Codex mode, you might have to be in a in a repo for this, but you'll see, like, the diffs of, the the sheet that it's creating and stuff like that and the file edits.
但在Work模式下，你是看不到这些的。
But in Work, you won't be able to see that.
我觉得这已经说得非常清楚了。
I think that's that's super clear.
另外，我还想深入了解下你们的生产力团队。
And then also the other thing I wanted to dive into was your the productivity team.
还有哪些其他内容？
What else is there?
首先，除了生产力团队之外，还有哪些主要的团队？
First of all, you know, what what what are the top level teams other than productivity?
难道生产力涵盖了所有方面吗？
Isn't productivity everything?
我们有一个专注于 ChatGPT 的团队，也就是面向消费者的核心聊天体验。我觉得这不能全算作生产力，因为人们每天用 ChatGPT 来搜索信息，比如怎么给亲人写消息，或者思考如何学习一个新话题等等。
You know, we have a team focused on on ChatGPT, like the the core chat experience for consumer, which is like, you know, not I think all productivity, like there's people are using ChatGPT every day for search to, you know, figure out how to write messages to loved ones, to think about how to, like, learn a new topic, etcetera.
所以还有更多功能，比如生成图片。
And so there's so much more inside to create images.
聊天功能里还有很多东西，你知道的，几亿用户都在用，这显然需要投入非常大的精力去专门做。
There's so much more in chat that, you know, the hundreds of millions of users are using that, you know, obviously, that that warrants, like, a a very dedicated effort.
还有专门负责企业版、基础设施和 API 之类的团队。
And there's teams focused on enterprise and infrastructure and API and stuff like that as well.
好的，对。
Will bring Yeah.
所以我两个都开着。
So I have them both running.
对。
Yeah.
这是工作。
This is Work.
这里有个Codex版本。
There's a Codex version here.
我选了5.6 SOL，所以得等一会儿。
I picked 5.6 SOL, so this will take a while.
嗯，明白了。
Uh-huh.
我觉得咱们就让它后台跑着，等它们跑完，再看看有什么不同。
I think I think we'll just keep it in the background, and, know, as as they finish, we'll look into some of the differences.
是的。
Yeah.
但如果你马上切换回Codex版本，就会发现
But immediately, I think if you flip back to the Codex version, you'll see that
它默认是基于 Git 的。
That it assumes it assumes Git.
对。
Yeah.
没错。
Exactly.
是的。
Yeah.
比如 Dynamic Island 就假定你是在一个 Git 仓库里操作。
Like, Dynamic Island assumes that you're in a Git repo.
你可能会漏掉一些内容，因为有些东西其实藏在思考的完整链条里。
And you might miss some stuff because some of it is, like, in the actual chain of thought.
哪些是改动的部分？我们该怎么显示它们？
What what were changes and how we display that?
但有没有哪项功能，你们原本想上线，结果收到反馈后决定放弃？
But Is is there an unintuitive like, is there a thing that you wanted to ship and then you got feedback and you're like, no.
咱们别这么做了。
Let's not do it.
这么做的理由是什么？
Like, what's the thinking behind that?
在 ChatGPT 工作流里？
In ChatGPT Work?
对。
Yeah.
我觉得我们本来可以走另一个方向，就是把体验完全隔开。
I think one direction we could have gone with this is, like, keeping the experiences, like, completely separate.
所以就是，为什么不用不同的应用？
So it's like, why Different apps.
没错。
Exactly.
比如用不同的应用，或者就算在同一个应用里，也提供完全不同的体验。
Like, different apps or even in the same app, like, different completely different experiences.
为什么要把它全都合并起来呢？
Like, why merge it all?
你知道吗，Codex 到底吸引人的是什么？
Like, what is, you know, Codex, obviously, people love?
为什么要把这些产品凑到一块儿？
Like, why why bring these products together?
我觉得这里的直觉是，我们的工作正随着人工智能发生巨大变化。
And I think the intuition here is that, like, all of our jobs are, like, changing dramatically with AI.
每隔几个月，我就感觉一觉醒来，做的事情跟几个月前完全不一样了。
Like, for, like, every few months, like, I feel like I wake up and I'm, like, doing a completely different thing than I was doing a few months ago.
我的假设是，好吧，
And my my hypothesis here is that alright.
我们假设，这项技术的一部分价值，就是给人赋能——比如帮你自动化那些枯燥的日常工作，让你能更快地提出更多想法，或者做点别的什么，现在你都能做到了。
So our hypothesis is that, like, part of what we're we're building in this this technology is giving people leverage, like, you know, the things maybe it's the more mundane parts of your job or or parts that, like, if you were able to automate, you'd able to share more ideas faster, or whatever, like, you're able to do now.
正因为这样，写代码的、写战略文档的、策划活动的、做市场推广的、做播客的，这些角色之间的界限，可能真的会慢慢模糊掉。
And because of that, like, that might actually blur the lines between someone who's like only writing code, or creating strategy docs, or, you know, planning events, or helping with marketing, or doing podcasts, or whatever.
对吧？
Right?
所以这些界限随着时间推移会慢慢模糊。
And so like, these things are gonna get blurred over time.
因此，想根据身份硬性划分界限，其实会很困难。
And so, like, trying to draw a hard boundary based on, like, the who you are is is gonna be is gonna be tough.
我们应该让用户有选择的空间，但别给他们设限。
And, we should enable users to choose, but we shouldn't box them in.
所以这里很多工作都集中在保持底层结构一致上。
And so a lot of the work that went in here, like, know, keeping the primitives the same.
比如插件功能，之所以在这款产品、ChatGPT 和云端保持统一，就是基于这个考虑。
Like, for example, plugins are, like, unified across this product and ChatGPT and the cloud was because of that.
我们的核心观点是，最终这些功能都会融合在一起。
It's this this thesis that, like, eventually things are gonna come together.
我们不想强加规定说什么时候该用哪个体验，但也不该把任何人框死。
And and we don't wanna be like, we wanna be prescriptive about when to be in neither experience, but we don't wanna box anyone in.
我不确定有没有用户还特别习惯旧版ChatGPT的使用方式，毕竟现在已经被Codex系统取代了。
I wonder if there's users who are very tuned to the old ChatGPT harness that is effectively now replaced by the the Codex harness.
我很难想象那种体验是什么样的，但也许他们更偏向于对话式的用法。
I can't imagine what that was, but maybe they're more the more conversational side.
你能对比一下这两个系统吗？
Can you con compare and contrast the the two harnesses?
因为只有你亲眼见过两者。
Because only you've seen it.
是的。
Yeah.
我的意思是，现有的ChatGPT系统其实到现在还一直存在。
I I mean, I think ChatGPT, the the existing harness, like, still exists today.
它现在依然在这款应用里运行着。
It, like, exists in this app.
就是那个经典版本。
The classic.
对吧？
Right?
你这个
The You
新开一个聊天窗口，就不会影响到工作了。
just start a new chat and you don't go under at work.
对吧？
Right?
是的。
Yeah.
如果你新开一个聊天窗口，进入对话，那就是在和ChatGPT交流，所以再试一次。
If you start a new chat and go to chat, then you're you're talking to ChatGPT, which so do another.
对。
Yeah.
但我猜，就某个实例来说。
But I guess, you know, on instance.
是的。
Yeah.
所以这个不会运行代码，
So this one's not gonna code,
或者它会
or it's gonna
按顺序执行。
be in line.
在沙盒里它并不按顺序。
It's not in line in a sandbox.
实际上，如果你在创建表格，系统会催你去工作。
Actually try to push you to to to go to work if you're creating a spreadsheet.
而这种方式，是路由的决定。
And this way, this is a router decision.
抱歉？
Sorry?
这是路由器的决定吗？
It's a router decision?
这是模型做出的判断，它发现你正在尝试做一些在‘工作模式’下处理会更合适的事情。
This is the decision that, you know, the model is making, and then, like, you know, it sees that you're able to or you're trying to do something that would be better served in work mode.
不过我觉得你的问题是，像 ChatGPT 或 Chat Harness 这样的聊天工具，它们到底有什么优势？
But I I think your question was, like, what what are the advantages of, like, the the chat, like ChatGPT, chat harness?
更广泛地说，我想基本做一期关于 Harness 工程实践的访谈记录。
It's more broadly, I wanna basically do an oral history of harness engineering.
对吧？
Right?
ChatGPT 的那个框架从 o1 时代一直用到现在，现在基本上要被 Codex 框架取代了。
The ChatGPT harness lasted us from, let's call it the o one era until now, and now it's being replaced by the Codex harness effectively.
它们俩现在有些重叠，但我挺好奇到底变了啥，如果真有变化的话。
And they're they're overlapping somewhat, but I'm curious what changed if, you know, if there is.
我的看法是，这就像是一个不断发散、收敛、再发散、再收敛的过程。
My perspective on this is, like, there's there's there's all there's sort of like a constant process of, like, divergence, convergence, divergence, convergence.
在聊天场景里，我之前提到的那些用例——比如搜索或者学习——我们其实一直在优化响应速度和个性体验。人们喜欢ChatGPT，正是因为这些方面我们长期投入、不断打磨。
And in chat, like, many of the use cases I was talking about before, like, you know, search or learning, I think we're we're really optimizing for latency and optimizing for personality and, like, different things that over time, like, the product the reason people love ChatGPT is because we've been optimizing for those things and working on them for so long.
至于Codex，我们发现，只要给AI代理一个像电脑这样无限灵活的环境，它就能完成非常强大的任务。
Codex, what we learned was that, like, if you give the agent access to this infinitely flexible environment as a computer, it can do really, really powerful things.
所以当我们考虑知识工作场景时，该选哪个模式呢？
And so when we think about, like, okay, well, for for knowledge work, like, what is which mode should we choose?
对我们来说，更自然的做法是把这种能力带到电脑环境中，同时把一些复杂的操作细节对普通用户隐藏起来，但依然让他们享受到同样的强大功能。
It was like it felt more natural to us to bring that to this, like, computer environment and, you know, maybe abstract some of the details of this computer away from users who might not be used to that, but, like, give them that same power.
但归根结底，我觉得所有场景都应该具备这种能力。
But ultimately, I think that we want the power in in all places.
对吧？
Right?
我们要做的，是到用户所在的地方去服务他们。
We wanna meet people where they are.
我确信，未来还会有大量工作要做，才能让所有场景下的表现都达到同等水平。
So I'm sure there'll be work down the road in order to to get things to be equivalently capable in in all scenarios.
但这只是取决于我们过去重点放在哪些产品上，现在又在关注什么。
But it's just a question of, like, what we've been focusing on the product on historically and what we're focusing on now.
我觉得除了什么时候用Codex、ChatGPT还是Work之外，你们新发布的模型也值得聊聊。
I think alongside that, outside of just harness and when to use Codex, ChatGPT, or Work, there's also the new models you've released.
对吧？
Right?
有什么建议吗？
Any guidance there?
大家总喜欢精打细算，比如高推理任务就用Terra，这种场景就用SOLE，别用别的。
So people love to min max what to use, like, only use Terra on high reasoning versus for this, you you wanna use SOLE here, ignore
一共有32种选择。
There's 32 options.
是的。
Yeah.
没错。
Yeah.
对。
Yeah.
不过话说回来，对于那些正在拓展、尝试用工具提升工作效率，却还不清楚这些技术具体怎么用的人，你有什么建议？
But that being said, know, for people that are expanding, so productivity, trying stuff for work that don't have the breakdown of what all this is, what's what's the advice?
对吧？
Right?
我觉得，在提建议之前，首先要明白，没有这些模型，这一切都不可能实现。
Well, I mean, I think before the advice, like, first thing is like, none of this would be possible without these models.
你之前不是问过，Work 的灵感来自哪里吗？
Like, I think you asked earlier, like, you know, what was like the inspiration for work?
早期我们看到 Codex 的表现，其实也是因为模型的能力在持续大幅提升。
Like, you know, early on, like, I I I mentioned like what we were seeing with Codex, that was also because the the models were getting infinitely more capable.
现在又在重演这一幕。
That's happening again.
我觉得，现在又是一次质的飞跃。
I think it's like another step function jump now.
关于设备端的问题，我们的默认设置一定要做到最好，我们对默认选项是有明确主张的。
And to answer the question on device, like, we want this default to be the best possible like, we wanna be opinionated about the default.
所以我们选了一个我们认为对所有人都最合适的默认配置。
And so we've we've chosen a default that we think is gonna be the best for everyone.
当然，对于高级用户，我们后台也提供了可调选项。
And, you know, we have, for power users, options under the hood.
有人可能会觉得现在选项太多了，我们正在努力简化它。
We could one could argue that there might be too many right now, and we're, you know, working on simplifying it.
但你可以自行调整推理层级，也可以根据需要切换不同的模型类型。
But you can extend, you know, the reasoning level, and you can change between the different model classes if you need to.
不过对大多数人来说，默认设置就是最合适的。
But the default should be the best for for most most use cases.
所以我建议大多数人就直接用默认设置就行。
So my advice to most people would be to stick to that.
如果你遇到某种情况，觉得效率或成本不够理想，或者智能表现不够好，那时候再考虑调整配置，看看能不能找到更好的方案。
And then, you know, if you reach a situation in which you think that you could you wanna try a different configuration if you're not seeing either the the efficiency on the on the cost side or or the the quality on the intelligence side, then you can change the defaults and see if you can get something better.
但我们觉得默认设置已经足够好了。
But but we think the the default should be good enough.
我有个想法想跟你聊聊，因为你比我有经验多了。
I have a I I'm just gonna run something by you since you you have way more experience than me.
我最近在用Soul Light，但加了目标导向。
I've recently been doing Soul Light, but with goal.
想法是目标能增强推理过程，只是会多一些终止点和轮次。
With the idea that the goal basically augments the reasoning effort, but with more terminations and turns.
对。
Mhmm.
这样理解对吗？
Is that a good way to think about it?
跟Soul Ultra或者Soul Extra High相比呢？
As opposed to Soul Ultra or Soul, you know, extra high.
是的。
Yeah.
这很难说，因为……
It's hard to say because Yeah.
这就像一种交互效应。
It's like an interaction effect.
没错。
Exactly.
这其实取决于个人偏好，比如你平时喜欢怎么和模型协作？
It's like there's a preference on you know, for you as an individual, like, how do you like to collaborate with the models?
你希望设置多少次这样的‘终止点’，好让你能随时介入，确保它在做正确的事？
Like, how many of those, like, terminations, as you call them, do you want where, you know, you can steer or make sure that it's doing the right thing?
我觉得大家最好试一试适合自己方式的方法。
I think generally, people should try whatever works for them.
我觉得，用Ultra或者多智能体系统，最适合那些特别复杂——比如开放式探索——或者特别容易陷入僵局的任务。
I think that, like, using Ultra or the, like, multi agent setups are best for, like, when you have, like, tasks that are either incredibly complicated, like open open expirations, or very paralyzable.
我觉得，即使是使用Goal模式，也更适合那些你能持续、可验证地推进的任务。
I think even for tasks using goal, I think, is best for for tasks that you know that you'll be able to make consistent progress in a way that's verifiable over time.
不过我觉得，大多数任务其实都不属于这两种情况。
But I think for most tasks, they actually don't fall into either of those buckets.
所以至少在刚开始的时候。
And so, like, at least when they're starting.
因此我觉得，最好的第一步就是先用默认配置试一下，然后看看接下来该怎么调整。
And so that's why I think the best first step is, like, trying it with the the default configuration, and then seeing, like, where you wanna go from there.
没错。
Right.
你们研究了
You guys worked on
一个滑块，这其实对减少恐慌感特别有帮助。
a slider, which actually is super helpful for reducing the amount of panic.
对。
Yeah.
是的。
Yeah.
在手机上用着还挺舒服的。
It's nice on mobile, at least.
这里有个不错的滑块。
There's a nice slider here.
我还没试过。
I haven't tried it.
你那里有高级视图，但如果你点了高级视图，对吧？
So you you you have you have the advanced view there, but if you click advanced view yeah?
对。
Yeah.
哦。
Oh.
就一个简单的滑块。
Just a nice slider.
对。
Yeah.
非常漂亮，色彩也很丰富。
Very very pretty, very colorful.
对。
Yeah.
那个想法是把这里的东西，就像，压缩成了一维，尽管实际上有多个维度。
That idea was here here was, like, reduced it to, one dimension, even though there's multiple dimensions.
对吧？
Right?
试着把它投影到单一维度上，方便用户理解。
Try to project it onto a single dimension for the user.
对。
Yeah.
比如，你知道的，一边代表速度和效率，另一边代表质量和细致程度。对。
Like, you know, something from that represents, like, know, speed and efficiency on one side Yeah.
然后另一边就是质量和细致程度。
And then, like, sort of, like, quality and thoroughness on the other side.
我有点不明白，为什么这里用‘Soul’这么多。
I am just puzzled that it uses Soul so much.
比如那个低阶的版本。
Like, the lower No.
我觉得蜘蛛，如果我没记错的话，应该是Terra。
I think spider, if I'm not mistaken, is Terra.
原来是这样。
Oh, it is.
是的。
Yeah.
这是个人化的设定。
It's e personal.
挺好的。
Nice.
所以他们把Terra默认只设成了光明版本。
So they they preset Terra to only be the light one.
展开剩余字幕（还有 480 条）
我明白了。
I see.
但我觉得其实很多人，应该更多人用Terra才对。
But, like, I think a lot of people actually would more people should use Terra.
第一，SoulKey的容量快撑不住了。
One, because SoulKey's running out of capacity.
这事儿跟我有关。
I'm the reason.
你懂的。
You know?
这是我们的退休计算器，花了十分钟。
Here's ten minutes of our retirement calculator.
哦，就是我们那个Excel表格啊。
Oh, that's the Excel thing we're doing.
天哪。
Oh my god.
你看，这算是工作，Codex 还在跑，咱们待会儿再继续。
Look This is work, and then Codex is still cooking, so we'll get back into it.
我觉得真正有意思的是，能看看他们的思考过程和推理方式。
I think it'll be interesting to actually see the thought process, the reasoning.
而且，说起来，这已经花了八分钟在工作上了。
And also, you know, I guess this is eight minutes on work.
Codex 还在运行中。
Codex is still cooking.
是的。
Yeah.
对了，我认识一个叫蔡嘉伟的人，你听说过吗？
And by the way, I so I've do you know Gabriel Chua?
他是 OpenAI 新加坡团队的成员。
He's part of the OpenAI Singapore team.
他给我看了这个，我简直震惊了，这看起来跟 Excel 一模一样。
He showed me this, and I was, like, pretty shocked that this looks like Excel.
对。
Yep.
它能直接编辑 Excel 文件。
It it edits Excel files.
你根本没买过 Excel 的授权。
You never paid an Excel license.
对吧？
Right?
但不知怎么的，这东西居然能用，简直就是智能 Excel。
Like but somehow this this is, like, kind of workable, and it's agentic Excel.
是的。
Yeah.
我们这次发布重点推的一个功能就是各种生成内容。
I mean, one of the big, like, pushes that we made for this launch was, like, artifacts.
对吧？
Right?
比如在模型这边，如果你拿这个和之前的5.5、5.4版本比，会发现这些生成内容的质量提升非常明显，产品层面也是同样。
Like, both on the on the model side, like, I think if you compare this with with five point five and and and five point four before that, you'll see that there's been pretty dramatic improvements in the quality of these artifacts, and then also on the product side.
用户体验方面也太夸张了。
The UX side is also crazy.
像托管网站这些功能，现在根本不用自己搭网页了。
Like, hosted sites and whatnot, no no longer needing to host your own little web page, like
我有个相关的故事。
Oh, I have a story about that.
我可以单独讲一下吗？
I can I can do a separate thing?
我得把这里的截图拿下来，不过咱们稍后再切到这部分。
I'll need to take the the the visuals here, but we'll we'll we'll cut to that later.
你们做这么大调整的时候有联合训练吗？因为你们和ChatGPT Work同一天发布5.6，是模型团队和工程团队有协作，还是只是巧合撞上了同一天？
Was there co training, I guess, because you are moving making this big move, and you launched five six on the same day as ChatGPT Work, was there influence between the model training teams and the harness teams, or do they did the launch dates just happen to line up with the same day?
我觉得我们和研究团队的合作非常紧密。
I think the you know, we collaborate heavily with the research teams.
我觉得这正是工作中最神奇、最有趣的部分之一。
And, I mean, I think that's like one of the most magical parts of the job, like the the most fun parts of the job.
但举个例子来说，你看到的这些成果，背后其实投入了大量工作，确保我们有合适的基础设施来训练模型，让它在这方面不断进步。
But, yeah, I mean, just just using artifacts as an example, like, you know, a lot of what you're seeing, like, underneath the hood, there's a lot of work that went into making sure that, like, you know, we had the right infra to be able to train the models to get better at this.
在产品层面，我们也打造了让用户能跟模型一起协作处理这类成果的流畅体验。
And then on the product side, like, had the right experience for users to be able to collaborate with the model on an artifact like this.
事实上，整个查看器的设计理念是：你可能根本不需要额外购买 Excel 许可证。
In fact, like, whole viewer, like, the intuition here is that, like, you know, it's not necessarily that you wouldn't need an Excel license.
这还只是第一阶段。
This is stage one.
对吧？
Right?
你当初做退休计算器的时候，应该不是想搞成这样的吧？
Like, is probably not what you meant when you're like making a retirement calculator.
你是想不断迭代的。
You wanna iterate.
而且，当你看到它的时候，如果它的保真度很高，就像你实际看到的，或者如果你把它发给Sean，你的同事们看到的那样。
And like, when you when you're seeing it and if this thing is high fidelity to like what you'd actually see in in or what your your coworkers would see if you were to send this to Sean.
我觉得，这让迭代变得容易多了，也让你对产品更有信心。
Like, that that, I think, makes it so easier and makes you trust the the the product in in terms of iteration.
当你说同事们会看到的时候，你是指会有多人、多团队的协作功能吗？
When you say coworkers would see, do you see a multiplayer, multi team collaboration with artifacts?
你们有考虑过这方面的事情吗？
Any any things you guys think about?
分享出去。
Share it.
对吧？
Right?
是的。
Yeah.
对。
Yeah.
这是我们目前正在积极考虑的事情。
It's it's something that, you know, we're actively thinking about.
另外，我们内部观察到一点，虽然不太方便透露具体规划，但很多时候有人来问我问题，我会先问 ChatGPT Work，然后再把答案回复给对方。
One thing that, you know, we've noticed internally without talking too much about the road map is that, like, I there's many times when someone will ping me about something, and I will ask ChatGPT Work the question, and then I'll ping them back the answer.
然后我就在想，要是我们三个人都在同一个共享环境里就好了。
And then I'll be thinking would be, you know, the three of us are just all on one hosted.
没错。
Exactly.
我会想，比如，你知道的，我是不是非得在这个流程里不可？
And I'll think about, like, you know, like, was I required in this loop?
或者，也许它被重新表述了，比如他们问的意思，或者是从某些背景信息里提取出来的，诸如此类。
Or and then maybe it was, you know, rephrased, like, what they were asking or pulled from certain context or whatever.
但是，你知道的，当我把答案给他们的时候，这个过程也是有信息损耗的。
But, like, you know, when I gave them back the answer, that process was also lossy.
对吧？
Right?
我只是把ChatGPT生成的内容，按我自己的理解转述给了他们。
Like, I gave them just, like, my interpretation of what ChatGPT Work cooked up.
但其实背后有很多上下文信息，比如在实际应用中的各种细节，这些都挺有意思的。
But, like, underneath the hood, there's so much context, like, in the rollout and stuff that that could be interesting.
所以你是说，答案其实是提前回复每一个收到的请求？
So, like, the answer was preemptively respond to every inbound request?
不是。
No.
这其实就是我工作中有时候会做的事。
And it's just, like, literally, like, this is what I do sometimes with my job.
我知道你就是复制粘贴，纯粹在AI和AI之间当个消息中转站。
I know you copy paste, and then you you know, you're just a message forwarding service from AI to AI.
但我觉得这挺有意思的。
But I think it's interesting.
对吧？
Right?
它能帮人们意识到，原来你能问AI这么多问题，还能交给它去办，很多人没试过根本想不到。
It helps people understand the capability of what you can ask and delegate that oftentimes people don't realize until they try or someone shows you, and then you're like, oh, okay.
明白了。
Okay.
原来如此。
I see.
是的。
Yeah.
我觉得还存在一个轻微的安全问题，你其实相当于权限控制层。
I think it's also there's also like a light security issue where, like, basically, you're the permissions layer.
没错，我本来也能查你查的所有东西，也能拿到自动回复，但也许我
Like, yes, I could query everything that you query, and I could get an automated response, but maybe I'm
不该看到这些。
not supposed to see it.
对。
Yeah.
而且确实是这样。
And that Yeah.
我根本不可能知道，因为我本来就不该知道那些我不知道的事。
There's no way I would know because I'm not supposed to know what I don't know.
特别是像 ChatGPT Work 这种，让你去连接插件，然后它会读取你本地的文件之类的。
Especially as, like, you know, with ChatGPT Work, are asking you to connect your plugins, and, you know, it's pulling from your local files and stuff like that.
助手能接触到的上下文信息量非常大，涉及很多个人隐私，这类东西我们必须得保护好。
Like, the the amount of context that the agent has access to is, like, deeply personal, and, like, that's something that, like, we need to preserve.
那肯定是个挑战。
So that'll be definitely a challenge.
有Excel，有PPT，还有Google Docs，你知道的，职场里的三大巨头。
There's Excel, there's PowerPoint, there's Docs, you know, the grand trio of work.
你觉得还有哪些其他的工作格式？
What other formats of work do you do you think about?
比如，你之前用过Airtable对吧。
You know, like, obviously, you worked on Airtable.
有没有可能未来会出现一个OpenAI版的Airtable？
Is there a future where there's, like, OpenAI Airtable?
我的意思是，如果你真要做出来，那会是个什么样子？
Like, know, like, what what what does that look like if if you ever ended up doing it?
这是个非常好的问题。
It's a really good question.
你没提到但我觉得很重要的一个是Sites，尤其是Sites的某一方面，大家在Twitter或者X上经常聊，把它当成一个原型工具。
I think I mean, one that you didn't bring up was Sites, and I think that was There's a part of this one one side of Sites that I think people commonly talk about, especially on Twitter and stuff, or X, of, like, you know, this sort of, like, prototyping tool.
实际上，这次发布我们就亲眼看到了这种情况。
And actually, like, we saw that happen with this launch even.
你们之前提到的那个模型滑块，几乎完全是用Sites做出来的。
The the model slider that you guys were referencing earlier, like, that was developed almost fully in a site.
设计、工程和产品团队之间的协作，就是在Sites上完成的，大家一边试用各种交互设计，一边感受体验，不断调整。
Like, you know, the the collaboration between design and engineering and product on that was, like, on a site where we play with, you know, the the the affordance and and and figure out how it feels and and all of that.
但另一个不太被讨论的方面，是Sites作为知识工作成果的载体。
But the other aspect that that I think is a little bit less talked about is, like, sites as, like, an an artifact for for knowledge work.
前几天我跟公司财务团队的人聊了聊，他们说现在团队每月做报告的时候，不再用PPT或者Excel了。
I was actually talking to someone the other day who was on, like, our our corporate finance team, and, like, they're mentioning how, like, now when they have these reports that they're they're working on as a team month to month, historically, things were in in slide decks and in spreadsheets.
现在全都放在Sites里了。
And now they're just in sites.
Sites成了他们团队协作的默认工具。
Like, Sites is the mechanism that they collaborate across the team.
原因在于，这种方式的信息传递效率更高。
And the reason is because it's, I think, it's, somewhat higher bandwidth.
像PPT和Excel确实很灵活，但用久了总会碰到瓶颈——要么你不知道怎么用某个功能，要么工具本身就不支持。
Like, you know, at these tools, like PowerPoint and Excel are, like, infinitely flexible, but at some point, you reach the boundary of, like, either as a human, you may not know how to use some feature or something, or the product itself doesn't support it.
但在Sites里，几乎想做什么都能做到。
But with a site, you kinda do anything.
你提出任何需求，基本都能实现。
You ask for anything, and you can get that.
一旦人们体验过这种便利，就会觉得特别有价值。
Once people see that magic, I think it's been really valuable.
对。
Yeah.
我来给你看看我的案例研究。
Let me show you my my case study.
这里面涉及所有热门话题，包括 ChatGPT Work，还有 5.6、token 十亿富翁、token 上限以及 Sites 和自动研究。
This involves all the hot topics, including ChatGPT Work, but also 5.6, token billionaires, and token maxing, and sites, and auto research.
我很喜欢一个叫 Strata 的游戏。
I'm a fan of this game called Strata.
它基本上就像个小桌游，你得用那种实体的积木块放在上面玩。
It's basically it's like a little board game that you that you play with physical blocks that come on top of it like that.
所以周末的时候，我拍了大概30张照片，直接丢给了ChatGPT。
So over the weekend, I I I took, like, 30 photos and just threw into ChatGPT.
处理了17亿个token之后，它生成了这样一个网站，里面有个完全能玩的游戏，支持3D积木摆放。因为需要实体积木，我还得找朋友来陪我练手，这样他们能变强，我才能跟他们切磋。
1,700,000,000 tokens later, out comes this site with a fully playable thing with with with three d block placement and everything because it requires physical blocks, and I needed friends to train on it so they can get better so I can play against them.
而且，我还能拿它来训练AI，这就是……
But also, I could also do things like train an AI on it, and that that's the
这是你的
That's your
自动研究。
auto research.
这就涉及自动研究了。
That gets into auto research.
所以你想训练自己的AI，然后让它们互相对弈。
So you want to train your own AIs and then make sure they self play against each other.
我得同时设置两个AI。
I need to set both AIs.
这是AI对AI，它们会自己对战。
So this is AI versus AI, and they're they're gonna self play.
显然，AI一开始很菜，你需要定义一个损失函数，让它不断进步。
Obviously, that the AI start out bad, and then you want to define a loss function and and get good.
我可不想亲自监督这一切。
I wasn't gonna supervise all this.
我当时在达拉斯和圣马特奥参加一个会议。
I was at I was at Dallas San Mateo attending a conference.
我最后做的是自动研究这个，还建立了一些基准测试，结果参数实在太多了，我根本看不过来。
What I ended up doing was auto researching in in on this and creating benchmarks, and that there was just way too many parameters for me to to read.
所以我开始让它帮我生成一个网站，它就做出了这个面板。
So I started asking it for a site, and it's created this this this this lab panel.
对于它生成的网站，有没有什么快捷方式或者快捷键？
Where is there is there a is there a shortcut for a site that is created?
你应该能在侧边栏里找到“站点”，就在侧边栏的顶部。
You should be able to go in the sidebar to sites, top of the sidebar.
就是左边那个侧边栏。
The left sidebar.
是这个吗？
This one?
哦，左边那个？
Oh, left?
是的
Yeah.
我直接滚到最顶端。
I just scroll all the way to the top.
哦。
Oh.
哦，上面写着‘站点’。
Oh, it says sites.
对。
Yeah.
在这儿呢
Oh, there
找到了。
you go.
对。
Yeah.
哦。
Oh.
所以它会自动生成这些网站。
So it create it creates the sites.
这其实不是我原本想要的，但我还是给你看看它弹出来的东西。
I don't I don't think this is a it is exactly what I what I wanted, but let me let me show you what it popped up.
对吧？
Right?
我觉得作为研究产物，清楚地传达具体做了什么非常重要。
Like, I think as a as a research artifact, it is very important to communicate exactly what is being done.
它输出了这个东西，后来我就开始发布它了。
Outputs this this thing which I eventually started publishing.
所以我后来把它从 Sites 上移走了，因为我想用比 Sites 更强大的数据库和基础设施，但这个东西确实是个可以动手试试的研究成果，让你去思考：你到底在训练 AI 时调了哪些超参数。
So I I moved it off of Sites because I wanted more database and infrastructure than Sites afforded me, but this is this is, like, a research output that you can start to mess with and, like, try to think about, like, what hyperparameters are you tuning for training your AIs.
我当时想搞清楚缩放规律，还做了各种游戏优化之类的东西。
And, like, I was trying to make, like, scaling laws and everything and doing all sorts of, like, game optimization stuff.
而且，你能把这东西直接当研究成果放出来，我就不需要再
And the fact that you can just kinda throw this up as a research artifact, like, I no longer need
读
to read
ChatGPT 的输出了。
ChatGPT output.
我看的是网站的输出。
I read site output.
但同时，这也带来了一大堆乱七八糟的东西。
But then there's also a huge sprawl.
你看，这东西有多长。
Like, look at how long this thing is.
里面数字太多了。
There's so many numbers.
真有点让人眼花缭乱。
It is pretty overwhelming.
所以我就得从那儿开始删减。
So then I have to start pruning from there.
但从 Markdown 转向这种形式，还挺有意思的。
But it's an interesting transition from markdown Yeah.
你输出的不再是简单的笔记，而是一个完整的网站。
Activity that you're putting out to yeah.
你输出的是一个功能齐全的网站。
You're putting out a whole functional site.
我觉得 Markdown 对读者来说并不是最优的选择。
Think markdown just isn't that optimal for people to read.
对吧？
Right?
干脆直接写个 HTML 网站算了，而且我觉得这方面的自定义空间其实很大，你说呢？
Might as well just write HTML website, and I don't know, I think you can do a lot with customizing this, right?
你有那些技能，能清楚地表达你想实现的效果。
You have your skills that explain what you want.
我注意到它们的表述太啰嗦了。
Like I noticed they're quite verbose.
我不需要这么多信息。
I don't need a lot of this information.
确实太啰嗦了。
It's very verbose.
有个网站并排着看的好处是，你可以不断调整，留下需要的，去掉多余的。
So, and then the nice thing of having a site side by side is, you know, you just iterate on what you want and what you don't.
对吧？
Right?
是的。
Yeah.
我不确定这会不会让你想起一些关于内部运作的故事。
I I don't know if any that triggers any stories for you of how it's run internally.
我这么做对吗？
Am I doing this right?
是的。
Yeah.
我的意思是，我们现在看到的这种工作流程，各种团队都在用，以前那种标准的交付物是PPT之类的，现在正逐渐变成网站。
I mean, I think that this is like a workflow that we're seeing, like, all different types of teams use, where, like, the the canonical artifact that was previously a deck or something is now becoming a site.
而因为网站就是HTML，所以它的灵活性几乎是无限的。
And, like, with a site, you because it's just HTML, you can, like, it's infinitely flexible.
如果你希望某个内容比在PPT里更突出——在PPT里可能被埋没了——现在你完全可以做到。
And so, you know, if you if you want to give more prominence to a certain thing that, like, in the slide deck would, you know, feel like it was buried, like, you can do that.
你可以把它做成首页的主视觉图。
You can have it be like the hero image.
对吧？
Right?
我觉得大家已经开始意识到这一点了。
And so I think that, like, people are starting to see that.
当然，要让这些网站更容易协作，还有不少工作要做。
There's obviously more work to be done to make these things, like, much more easier easy to collaborate on.
你提到它们又长又啰嗦，其实可以拆开。
You mentioned that they're very they're long and verbose, could be broken up.
我肯定
I'm sure that
有个特别长的。
there's something Super long.
是的。
Yeah.
对。
Yeah.
但我认为我们正开始意识到，这种格式其实特别灵活，比以前任何方式都强，挺有意思的。
But I I think we're starting to see that, like, there there is this aspect of this is a really interesting format for for people to use that's, like, much more flexible than than what they ever had before.
我觉得你的工作本身也变得有点元了。
I think your job also comes becomes kinda meta.
你不是在设计产品，而是在设计一个用来制作产品的工具。
You're not designing the products, you're designing a product to make products.
我很想知道你是怎么处理这个问题的。
And I'm curious how you manage that.
我觉得我们在看用户体验时，一直在思考一个问题：怎么在简洁和功能之间取得平衡？
I I think one thing that we've been like, when we look at the UX, like, that that we've been thinking a lot about is how can we balance, like, simplicity with capability?
就像你所说的，我们设计的产品是用来打造其他东西的。
Like, if we if we're designing a product, like you said, that, like, is is made to make up build other things.
对吧？
Right?
你可以用它做出各种各样的东西。
You can build so many different things.
但要是把所有功能都堆在你面前，你肯定会觉得不知所措。
But we can't put that all in front of you because you'll get overwhelmed.
没错。
Yes.
我们之前在和ChatGPT打交道时，也遇到过类似的问题和挑战。
And so we had similar problem or or similar challenges even chat with ChatGPT.
但现在尤其如此，能做的事情太多了，我们一直在努力平衡的是：如何给用户提供足够丰富的界面，让他们能自由表达需求，确认代理用的是正确的工具、调用的是正确的数据源等等。
But especially now, like, when there's so much that can be done, I think the balance that we're constantly trying to strike is, like, how can we give the user enough of a UI surface where, you know, they can be expressive, they can they can tell the the agent what they need, they can verify that it's using the right tools, it's pulling from the right sources, etcetera.
但同时又不会干扰他们。
But then it gets out of the way.
我们该怎么构建一个系统，让它能主动展示给用户能做什么，而不是靠我们去告诉他们？
And then how can we build the right system such that we can show them instead of telling them what can be done?
因为很多情况下，用户是怎么发现下一个使用场景的，这才是关键。
Because so much of this is gonna be like, how do they discover the next use case?
再之后的下一个场景呢？如果他们真想让AI彻底赋能自己的话。
And the next one after that, if they really want to, you know, to to be super powered by AI.
是的。
Yeah.
这挺有意思的。
It's interesting.
我觉得每个人的做法都不太一样。
I feel like everyone also just has a different way to do it.
对吧？
Right?
我之前也做过一个类似的版本，同一个游戏。
I made a similar version of this, same game.
我连棋盘或规则都没拍过照。
I didn't take any pictures of board or rule game.
我扔了个目标，十八分五十三秒后才搞定。
I threw an at goal eighteen minutes fifty three seconds later.
经过大量token的尝试，我最终也做出了一个类似的版本。
A lot of tokens later, I've got a similar version.
当然，没用上那些自动研究之类的高级功能，不过——
Obviously, not with all the auto research and whatnot, but,
你知道的，你得跟上最新的潮流。
you know You gotta do all the latest trends.
而且，没错，我是用Codex做的，虽然没成功，但挺有意思的。
And, yeah, I did it with did it with Codex, not work, but it's interesting.
对吧？
Right?
对。
Yeah.
这明显是 GPT 生成的头像。
And this is obviously GPT image generating the pro avatars.
对游戏设计很有帮助。
Very good for game design.
比如，一个
Like, a
很多游戏设计师当时都特别热衷于用 GPT 来生成图片。
lot of game designers were, like, really into GPT image.
我得说，更广泛的结论大概是，我们做这些其实主要是为了测试这些工具。
I will say, like, the the broader takeaway probably is the reason that we do this is more so just to test the tools.
对吧？
Right?
比如，这也是为了测试 5.6 版本刚出来时的效果。
Like, this was also a test for 5.6 came out.
我之前那款游戏是在 5.5 版本上做的。
I had done the game on 5.5.
对吧？
Right?
现在我不再需要它了，因为我已经突破了那些限制。
The ability for me to no longer need it to I had defeated the rules.
这是一款相当小众的游戏。
It's it's a pretty niche game.
它自己搞不定这个。
It couldn't find how to do this on its own.
对，没错。
Oh, yeah.
5.6 它
5.6 It
就是自动分配。
is auto distribution.
所以我才特别想试试5.6的这个功能。
That's why I was also very keen on testing the 5.6 capability.
不过你知道，随着工作推进和新东西出来，这些只是我们用来测试的一些方法。
But, you know, this is just as as work comes out, as new things come out, these are just our side ways to test things.
对吧？
Right?
对。
Yeah.
大概是某种内部评估吧。
It's some kind of private eval, I guess.
是的。
Yeah.
并不是所有东西都是私密的。
Is not all this private.
但也很有价值，因为你现在可以把这个分享给朋友。
But also valuable, because now you can send this to your friends.
我是因为看到这个才了解到这个游戏的。
And I mean, I learned about this game through seeing this.
这是一款
It's a
很难的游戏。
hard game.
他水平很高。
He's very good.
没人跟你竞争的时候，反而更舒服。
It's good to when no one is competing with you.
不过，这确实是个经典的强化学习问题，就是通过自我对弈来训练游戏AI。
But, yes, it's a classic RL problem of, like, self play, bootstrapping your game AI.
对。
Yeah.
你有没有发现，工作很容易变成个人的事，而个人的事又很容易变成工作？
You see how easily work becomes personal, and personal becomes work?
因为我做的一些个人项目，后来直接影响了同事，因为我给他们看了，他们就说，哦，原来GPT还能这么用，我觉得这其实是一种增长策略。
Because the thing I do for personal, it actually directly informs people I work with because I showed it to them, and they were like, oh, you can do that with GPT, which I imagine is a growth strategy.
是的。
Yeah.
‘少说多展示’这一点特别关键，我觉得我们到现在还没完全搞明白，就是该多让别人亲眼看到产品能做什么，而不是光靠文章、引导教程这些方式去教他们。
The the show not tell is a big piece that, you know, I think we've we're not still not fully cracked of, like, you know, showing people all the the things that they can do with the product versus, like, trying to teach that to them through, like, you know, articles or onboarding or whatever.
对。
Yeah.
在关键时刻直接展示给他们看。
Submitting them in the moment.
这对我来说是有职业风险的，因为我以前做开发者关系，那工作就是得做示范。
It's a career risk for me because I used to be in developer relations, right, where your job is to show.
然后你就说，你什么意思？
And then you're like, what do you mean?
其实你不需要那样，你的工作是去讲解。
You you don't you don't need actually, your job is to tell.
对。
Mhmm.
然后产品那边的人会说，哎，如果我们的产品足够直观，那就不需要你了。
And then and then but the product people are like, well, we don't need you if our product is intuitive enough.
所以
So
对。
Yeah.
我的意思是，这就是模型的魅力所在。
I mean, that's the magic of the the models.
所以你可以把讲解或演示的内容，量身定制成用户真正需要的样子，比如他们关心的点、过去的操作习惯，以及他们目前在使用产品时处于哪个阶段。
So you can tailor the telling or the showing to, like, specifically what the user needs, like, what what they care about, what they've done in the past, exactly where they are on the adoption journey.
所以我觉得这是一个非常大的机会。
So I think that's, like, gonna be a super big opportunity.
现在要定制化地展示内容，好像越来越容易了。
Seems easier and easier now to tailor custom showing.
对吧？
Right?
每个人的需求都不一样。
People have different use cases.
你说过不想把不同的人硬性分到不同类别里，
As much as you said you don't wanna segment different people into different buckets.
对吧？
Right?
但对不同类别的用户来说，其实也不难做到差异化展示。
It's also not that hard too for people that are in different categories.
不过我想问一下，你之前提到你们团队更关注的是什么来着？
But the the question, I guess, is you said your team is more broadly on what was the term you used?
生产力？
Productivity?
生产力。
Productivity.
对。
Yeah.
生产力。
Productivity.
其实就是现在的工作内容。
Which is now work, basically.
是工作吗？
Is it work?
还有没有我们没覆盖到的其他使用场景？
Is there another distribution that we're not hitting?
有没有一群人，他们的需求跟ChatGPT、Codex或者工作工具都不一样？
Is there a group of people that will have something different than ChatGPT, Codex, or Work?
大众没关注的领域，是不是还有更多空间？
Is there is there more that the mass isn't targeting?
我觉得这更像是一个逐步推进的过程。
I see it as like a sequencing.
也就是说，我们的愿景是让每个普通人都能用上实用的智能助手。
Like, you know, the the vision is, like, bring useful agents to everyone.
我们一开始是从开发者入手的。
We started with, like, developers.
历史上，开发者一向都是最早尝试新东西的那批人。
Like, developers historically are, like, early adopters.
他们愿意忍受更多的麻烦，自己动手配置各种东西等等。
They're willing to put up with more friction, set things up, etcetera.
Codex 就是从这个群体开始的。
Like, that's where, you know, Codex started.
我觉得下一个机会，是所谓的通用知识型工作，也就是开发者周边的那些其他岗位。
I think the next opportunity is, like, sort of what we call general knowledge work, you know, all the other functions around developers.
从开发者转向这个新群体时，显然会遇到一些挑战，比如我们之前说的‘少说多做’——得让产品更直观易懂，还得加入一些对这群人更重要的新功能，像生成成果、操作电脑这些，以前主要是为开发者设计的。
I think when when you go from developers to this segment, like, there's inherent challenges, obviously, with, like, you know, this this show not tell thing that we're talking about, making the product more understandable, bringing in new capabilities that matter more for for this cohort, that matter for developers, things like artifacts, things like computer use, etcetera.
然后我觉得，同样的经验也可以用在一般知识工作者身上，就像我们当初从开发者身上学到的那样。
And then I think, like, the the same learnings, like, similarly how we took the learnings from developers and brought it to, you know, general knowledge work.
下一步，就是把从一般知识工作者那里学到的经验，推广到每个人身上，不管他们平时在做什么。
Next The stage will be like taking the learnings from general knowledge work and bringing it to everyone no matter what they're doing in their lives.
其实我们已经看到一些苗头了。
And we're already seeing that a little bit.
比如你提到的这个小游戏，它正好处在娱乐和个人生活与专业工作之间的交界地带。
Like, this this game example that you have is, you know, something that's like on the border of like fun and personal life to to, you know, your your professional life.
我每天在家全天候用ChatGPT处理所有事情，不管做什么都靠它。
I use ChatGPT Work full time at home for everything, like, for for whatever I'm doing.
前几天我还用它帮我规划了一周的菜单，顺便存到它的环境里，方便我随时回头查看。
I used it the other day to come up with a meal plan and, like, you know, save that on on the, like, computer environment that it has and something that I can continue going back to.
现在每个人都在这么做吗？
Like, is everyone doing that yet?
应该还没有，毕竟这东西还在完善中。
Probably not because the thing says work on it.
但最终，我们还是希望能让每个人都达到那个状态。
But eventually, you know, we wanna get people there.
生活全面智能化。
Tragibuty life.
是的。
Yeah.
没错。
Exactly.
烹饪全面智能化。
Tragibuty cooking.
但我认为这里头蕴藏着巨大的机会。
But I think there's a lot of there's a lot of opportunity there.
我觉得，我们先在软件工程领域打下了基础，接下来要把从软件工程中学到的经验，应用到知识工作，再推广到每个人的生活里。
But I I see it as, you know, we're we're built we built the foundation in in in software engineering, and we're gonna take the same learnings that we take from software engineering to knowledge work, knowledge work to to everyone.
你有什么给高级用户的小贴士吗？
Do have any power user advice?
我觉得有一群人会完全沉浸其中，把这用在方方面面，24小时不停歇。
I feel like there's a group of people that will live it, use it for everything, stay on it twenty four seven.
对。
Yeah.
但在这群人和另一些人之间，还存在一段差距——那些人呢，就是觉得这东西用在工作上还行，偶尔用一用，有时候提个问题而已。
And then there's a bit of a gap between that crew and people that, know, okay, I use it for work, I use it occasionally, sometimes I pipe questions.
你有什么建议吗？有没有什么经验或心得，能帮我们弥合这个差距？
Any advice, any learnings, anything you recommend, or just, you know, takeaways that you found that help bridge that gap?
我觉得有几点我观察到的，首先是，拓宽你对可能性的想象特别重要。
I think a couple of things that I've seen is like, one, that it really helps to broaden your imagination of what's possible.
这对我来说也是一个新的领悟。
And this has been a learning even for me.
你知道，技术进步得太快了，就在三个月前，我们还觉得这些模型根本不可能做到这些事。
Like, you know, the technology has progressed so fast that, you know, it was something that, like, even three months ago, was like, no way no way that the models can do this.
现在呢，真是让人惊叹。
Like, now it's like, wow.
它真的能做到。
It's like, it actually can.
举个例子吧。
Like Give an example.
我们现在正经历内部的评审周期。
We're going through right now there are, like, review cycle internally.
大家总说，模型在这方面特别拿手。
And, you know, people always talked about this as like kind of a a thing that the the models are good at.
老生常谈了，反正没人愿意写评审，干脆让AI代劳。
Like, you know, there's a cliche of like, okay, like, no one wants to be writing reviews, and, like, we just use AI to do it.
但说实话，
But I mean,
不过说正经的，
and then In all seriousness.
也要自己来评估一下。
Evaluate it as well.
是的
Yeah.
没错
Exactly.
说真的，以前这事儿 basically 就是个形式而已
In all seriousness before, it was like just like slot basically.
我觉得有点帮助，但说实话效率并不高
And, like, I I think it was helpful, but, you know, not super productive.
现在我发现，这模型能比我做得好得多，尤其是在这种能整合上下文的环境里——比如了解每个人在忙什么、他们做过哪些有影响力的事、突出那些我可能根本没注意到的成就，因为它能接触到所有信息
Now I've found out, like, the model can do a much, much better job of me, especially in this environment of, pulling context on, like, what people are up to, how they've like the things that they've done to make a difference, highlighting, like, you know, wins that they've had that, like, I might may not even have seen, you know, has access to, like, everything.
对吧
Right?
比如代码、他们发现的问题、代码评审、Slack 上的对话，全都涵盖
Like, the code, like, you know, things that they've caught, reviews, Slack, everything.
所以在这一块，它的能力真的特别强
And so it's, like, incredibly powerful in that domain.
六个月前，我们刚做完上一轮。
And, like, just, like, six months ago was the last time we did this cycle.
我当时试过用，但根本没什么用。
Like, I didn't even I tried using it, but it was not at all helpful.
这次却特别有帮助。
And this time, it's been, like, incredibly helpful.
我觉得，哪怕之前试过的东西，也别轻易放弃，继续探索还能做到什么，这可能是我最重要的建议。
And, like, so I think continuing to push the the frontier of imagination of what's possible, even if you tried something before, I think, is maybe the my biggest piece of advice.
还有就是，你投入得越多，尤其是在这种环境下——模型能访问你电脑上或ChatGPT工作区里的所有内容——你就能慢慢积累自己的资料库，模型也会一直保留这些信息。
The other, I guess, thing is, like, the more the you more put in, especially in this environment where, like, you know, the model has access to to everything on your on your computer or in ChatGPT Work, like, you can create, you know, artifacts over time and save them in your library, like, the model will continue having access to those.
你给它的信息越丰富，不管是生活还是工作相关的，它的价值就越高。
The Like, more information you give it about whatever domain you're in, whether it's your life or your work, the more valuable it becomes.
它带来的帮助可能会让你意想不到。
And it'll become both valuable in ways that might surprise you.
它甚至可能主动从你都没想过的角度调用上下文，但前提是，它得先接触到这些信息和工具。
It might pull from context in a way that may be proactive and that you might not even have thought about, but it needs to have access to those to that those tools or that context first.
我总是想聊聊评价这件事，因为这真的挺敏感的。
One thing I always just wanna talk about the review stuff because I still that's a very sensitive thing.
你是创始人，带过团队，也招过人。
And you're you're a founder, you've managed people, you've hired people.
我自己当经理，特别不敢用AI生成的内容来评价人，感觉像是不把人放在心上。
As manager myself, I'm very reticent to put out any LLM generated things, especially when it comes to people, because it feels like you don't care.
对。
Mhmm.
在OpenAI，大家显然更接受让GPT来打分。
Presumably at OpenAI, people are obviously more open to being basically rated by GBT.
但这方面有没有什么不成文的规矩？
But are there any unofficial rules around this?
比如，大家平时是怎么把握分寸的？
Like, what's the etiquette?
我觉得规矩就是，我绝不会只靠AI写完一份评价，然后直接拿去给人看。
Oh, I mean, I think the etiquette is that, like, I wouldn't ever write something via, like, only solely via AI and, like, present it as, like, a review for someone.
我刚才说的更多是收集背景信息。
What I was talking about is more like gathering context.
对。
Yeah.
关键在于
That's the place where
其实就是搜索。
it's So it's just search.
这是一种有主动性的搜索。
It's agentic search.
是的。
Yeah.
这属于有主动性的搜索，但你能更灵活地调整和引导它，比以前强多了。
It's like agentic search, but, you know, that you can tailor and steer much more capably than you could before.
而且因为现在所有东西都连在一起了，形成了一种良性循环。
And because like the the thing is it's all there's sort of a flywheel happening.
对吧？
Right?
因为有了Codex，人们现在能做的事情比以往多得多，ChatGPT Work也是一样。
Because of Codex, people are able to do, and because of ChatGPT Work, people are able to do so much more now than ever before.
如果你能做更多事情，也更容易忽略一些东西。
And if you're able to do so much more, it's easy to miss things as well.
所以我们得用这些同样的工具，跟上大家带来的各种影响，弄清楚我们能在哪些地方提供帮助。
And so, like, I think we need to use these same tools to keep up with all the impact that people are having and and and understand, you know, where where we can be helpful.
我觉得，说实话，我经营一家小公司，搜索起来很容易。
I think the the the thing, like, obviously, I I run a small company, so easy to search.
但在OpenAI这种规模，你们在Slack里发那么多消息，你觉得会不会漏掉什么？
But at the scale of OpenAI, with the amount of messages that you guys put in Slack, do you think that it misses things?
很可能漏掉，但我自己也会漏掉一些事情。
Probably, but I think that I also miss things.
其实有没有漏掉都无所谓。
Like, it doesn't matter.
我觉得
I think
有时候吧。
sometimes.
得达到人类的水平。
It needs to be human level.
这都是相对的。
It's all relative.
对吧？
Right?
是的。
Yeah.
有时候它能发现你想不到的东西，还挺好的。
Sometimes it's nice when it finds things you wouldn't.
对吧？
Right?
比如现在，我的Codex系统设置是这样的：每个项目都有一个独立的笔记文件，它会自动把学习心得写进去。
Like, now, my Codex system prompts, they're set up in such a way that every project I have has a separate notes MD, and it just writes learnings to there.
然后全局的笔记可以从所有这些文件里提取内容。
And then the the global one can pull from all these.
有时候它会突然提醒我，哦，你四个月前做的那个项目。
So sometimes it'll be like, oh, there's this project you did like four months ago.
这是当时记下的一个笔记，它就随机把这段内容重新拉回到当前上下文中。
Here's a note that we had, and it randomly pulls it back into context.
这种事我根本不会主动去做。
I would never do.
我压根儿就没想过这回事。
I haven't thought about it.
然后我就说，好吧。
And I'm like, okay.
这简直超出了人类的能力范围。
This is quite superhuman.
对吧？
Right?
比如，它能省下好多时间，不用再反复整理或找那些早就做过的资料。
Like, stuff that would and, you know, it'll save, like, hours on chunking of stuff or find something that's already been done.
虽然它偶尔会漏掉一些东西，但我自己也常漏，不过一旦它找到了，真的特别有用。
I'm like, as much as it might miss stuff, I would too, but it's very useful when it finds stuff.
我其实有个挺简单的办法来解决这个问题。
And I have like a very, you know, non super engineered solution to this.
就是用 Markdown 文件，需要的时候自动调出来。
It's just markdown files that get pulled whenever they want.
是的。
Yeah.
我倒有个挺有趣的小故事想说说。
I actually have a funny anecdote about this.
最近准备上线的时候，团队已经连续几个月全力投入了。
Like, recently, gearing up to this launch, you know, the team has been, you know, really cooking on it for for for a couple months.
那段时间里，Slack、文档和其他地方的聊天信息多得不得了。
And over that time, like, there's so much conversation chatter going on in Slack and Docs and elsewhere.
团队里有个人设了个自动任务，定时扫描所有动态，挑出最搞笑的梗，然后发到我们的共享频道里。
And one one of the members of the team set up this scheduled tasks, like automation, to, like, look at everything that's going on and, like, come up with the best memes, and then post it in one of our shared channels.
这件事有两个特别有意思的地方。
And, like, there are two cool things about this.
首先，我觉得这些模型随着时间推移，真的开始变得好笑了。
Like, the first is, I think the models are, you know, over time, like, actually starting to become, like Funny.
挺逗的。
Funny.
挺好的。
Nice.
但一年前，根本不是这样。
Whereas, like, you know, a year ago, like, that was not at all the case.
其次，这正好印证了你之前说的。
The second is that it was what you were saying.
它们会以一些你根本想不到的方式发现东西，还能建立起你平时根本不会想到的关联。
Like, they find things that in surprising ways that you may not have not have thought of and, like, create connections that you may not have thought of.
这其实对生成表情包特别有帮助，因为你能看到一些真正让你惊讶的东西，而这种惊讶恰恰让人觉得好笑。
And that really helps with like the meme generation because then you can see something that, you know, genuinely surprises you and then and is funny in that way.
没错。
So, yeah.
当然，这肯定不是用这项技术最高效的方式，但它确实展现了这种新兴能力——能发现那些你原本根本无从得知的信息。
Mean, obviously, that's like not like the most productive use of this the the technology, but it it does it doesn't cover this, like, this capability that's emerging, which is just like defined information that you otherwise would not know of.
说到发布，我觉得这次可能是我见过最成功的一次发布。
Talking about the the launch, I I think I have pretty much said this is the most successful launch in a long time.
我个人觉得甚至比5.0版本还要成功，而且你们还宣布了有一千万用户。
I think even more successful personally than five point o, and you're announcing 10,000,000 users.
感觉有什么不一样吗？
Does it feel different?
你经历过那么多产品发布了。
You've been through a lot of launches.
我觉得这像是一个里程碑。
I think it feels like a culmination.
嗯，我觉得有两点。
Well, I I think two things.
第一，这确实是个终点，就像我之前说的，我们长期追求的这个愿景和使命终于实现了。
One, it feels like a culmination, like I was mentioning earlier, like this, like, vision mission that we've been on for a long time.
我之前说过，我们内部早就见识过Codex的神奇之处，现在能把它带给更多人，看到它真正发挥作用，达到你提到的用户规模，真的让人特别兴奋。
Like I said, we saw the magic of Codex internally, and then we're, like, extremely excited to bring this to many more people and to see it working, to, like, see us reach, you know, the distribution goal I mean, numbers that you mentioned.
我觉得这太了不起了，简直让人热血沸腾。
Like, I think that's, like, huge and and and super exciting.
但另一方面，还有太多事情等着我们去做。
The flip side of that is, like, there's so much more to do too.
这一点也同样让人激动不已。
Like, that's also really exciting.
你知道，整个ChatGPT这个产品，几乎成了人工智能的代名词，深受大家喜爱，用户数都达到数亿了。
Like, you know, ChatGPT as a whole, like, the the this product that, you know, everyone almost equates to AI and, like, loves, you know, has hundreds of millions of users.
一千万确实很了不起，但我们得让每个人都用上。
And so, like, 10,000,000 is really cool, but, like, we we need to get this to everyone.
每个人都该感受到这种神奇。
Like, we need everyone to feel this magic.
所以，接下来我们要走的路就是这个。
And so that's the next step from here.
不过，我对到现在为止的进展和未来的可能性真的特别兴奋。
But, yeah, I think extremely pumped about how how it's going so far and and the opportunities.
太棒了。
Awesome.
我其实还想说一下，因为我一直密切关注着这个数字。
I did want to also because I've I've I've been tracking the the number closely.
某个时候，用户群体已经从单纯的Codex用户，变成了Codex加ChatGPT Work，毕竟底层是一套系统。
It transitioned at some point from just Codex users to Codex plus ChatGPT Work, obviously, because it's the same harness.
关键就在于，你根本没法把这两部分分开算。
The whole point is that you don't you can you can't count them separately.
你们大概有十亿个 ChatGPT 用户吗？
Do you have roughly a billion ChatGPT users?
是直接一下子跳到十亿吗？
What is it just jump to 1,000,000,000 right away?
ChatGPT 默认就是十亿用户吗？还是说不是？
Like, isn't that the default on ChatGPT or no?
如果你已经在用 ChatGPT，我们不会默认把你转到 ChatGPT Work 里。
We don't default you into ChatGPT Work if you're on ChatGPT.
如果你有空的话。
If you're free.
对。
Yeah.
而且目前只有付费用户才能使用。
It's also only available to to paid users right now.
接下来我觉得得有个过程，就是教育用户了解这个产品的价值，让他们去试用，根据反馈来不断优化，慢慢把它做得更好。
And and then I think there's, like, a process of, you know, educating users of what is the value of this product, having them try learning from their feedback, and making it better over time.
但我们的目标其实是让现在喜欢ChatGPT的用户，真正感受到ChatGPT Work的强大之处。
But, I mean, the goal is to, you know, get as many of of the people who who love ChatGPT today to, like, feel the feel the power of ChatGPT Work.
但我认为这会是一个逐步的过程。
But I think it'll be a journey.
是的。
Yeah.
在可预见的未来，Codex这个品牌还会继续存在。
Codex will still be alive as a brand for the foreseeable future.
对。
Yeah.
我们会根据界面需求，在两者之间灵活切换。
And we'll just toggle between them as as as needed for UI stuff.
是的。
Yeah.
我觉得这个观点甚至比你说的还要更关键。
I think it's an even stronger point than that.
我们的本意就是，开发者一直以来都是我们的核心用户群体，我们会继续这样对待他们。
Like, I think we fully intend to like, you know, treat developer like, developers have been know, a core market for us for so long.
在软件开发方面，我们还有很多可以改进的地方，让 Codex 更加出色，这些工作我们会持续下去。
And, like, there's there's so much more that we can do to make Codex great specifically for software development, and we'll continue to do that.
这完全不会削弱这一点。
This doesn't take away from that at all.
相反，它反而能提升 Codex 这类工具的实用性，因为你现在可以无缝地在编写变更、生成代码产物，或者进行重构搜索之间切换。
If anything, it should increase the utility of something like Codex, because now you can move seamlessly between writing a diff to creating an artifact or, you know Yeah.
比如搜索重构方案。
Doing a search for refactor.
我有点好奇，这种术语会不会也影响到非技术人员。
I do wonder how much this terminology leaks to the nontenical user.
如果我要用产物功能，他们是不是也得学会说‘产物’这个词？
Like, do they have to learn to say artifacts if I want artifacts?
你知道的
Or, you know
挺有意思的。
It's funny.
我们内部叫它“产物”，因为团队内部都是这么叫的。
Like, we call artifacts internally because that's what the teams call it internally.
反正没人会这么说。
Like, no one says that.
没人会管它叫“产物”。
No one calls it an artifact.
不过我觉得大家通常会习惯用他们熟悉的方式来描述事物。
I but I think that people, like, often, like, describe things whatever they're used to.
对吧？
Right?
所以如果 ChatGPT Work 擅长做幻灯片，他们就会说它擅长做幻灯片，而这正是我们想要的效果。
So if, you know, ChatGPT Work is good at creating slides, they'll say ChatGPT Work is good at creating slides, and that's actually what we want.
还有一个大点，我是说，现在都2026年7月了。
One big another I mean, it's July 2026.
还有一个很重要的事，就是OpenAI的OpenClaw，我觉得很多人第一次真正把智能体用在个人事务上，同时也悄悄渗透到了工作中。
One big thing that also happens in for OpenAI was OpenClaw, and that's, I think, a lot of people's first time really maxing an agent for personal stuff, but also crossing over to work in in a some same way.
据我所知，OpenClaw到现在还是独立运作的。
As far as as far as I understand, OpenClaw is still independent.
那你有没有经历过自己的OpenClaw时刻？
But did you go through your own OpenClaw moments?
有没有从OpenClaw中学到什么，然后用到Codex上，或者反过来？
Were there any lessons that you took from OpenClaw to Codex or back?
随便吧。
Whatever.
我觉得这里面有很多启发。
I think there's a lot of inspiration.
我确实也经历过自己的OpenClaw时刻。
I I did go through my own OpenClaw moment.
对。
Yeah.
说说看。
Tell the story.
我和我老婆当时弄了个 OpenClaw，想用它来打理家里的大小事务。
Me and my my my wife, like, set up an OpenClaw to, like, try to manage everything in our house.
虽然家里的事儿不算特别多，但用起来还挺管用的。
Not that there's, like, a ton, but it was, like, actually quite useful.
我们给它接入了日历，让它帮我们创建日程安排之类的。
We gave it a calendar and started, you know, creating events for us and stuff.
后来那台笔记本电脑坏了，之后也就没再有机会把它捡起来继续用。
At some point, the the laptop they were running on had died, and then never got a chance to to pick it back up.
不过确实有很多值得借鉴的地方。
But there's a lot of inspiration there.
比如 ChatGPT 在网页和移动端的工作模式，它能提供一个持久的电脑环境，你可以存文件，而且这些文件在多次会话之间都能保留下来。
Like, you know, in ChatGPT Work in web and mobile, like, you you get access to, like, persistent computer environment where, you know, you can store files, and those files stay around between sessions.
这个想法就是为了支持这类使用场景。
And the idea is to be able to enable use cases like this.
我们团队里有个人现在用ChatGPT Work来替代以前的OpenClaw，我觉得这完全过渡过来了，主要是用来安排锻炼计划和记录饮食，说到底还是个人事务。
One of the members of our team actually uses ChatGPT Work for what they used OpenClaw from before, and I feel like it has, like, completely transitioned, which is, like, workout planning and, like, meal tracking, which, again, it's like a work y thing.
对吧？
Right?
严格来说不算工作，但属于个人效率管理的范畴。
It's, like, not work necessarily, but it's, like, in personal productivity space.
但它具备了所有相同的基础功能。
But it has all the same primitives.
比如可以安排定时任务。
So it has scheduled tasks.
还能在文件系统里存储文件，并且能跨会话调用这些文件。
It has the ability to store files on a file system, it has the ability to like reference those things over time.
于是你就会看到类似的应用场景不断出现，这真的挺有意思的。
And so you start to see the same types of use cases emerge, which has been really cool.
ChatGPT Work有没有可能彻底取代OpenClaw？
Is there a point that ChatGPT Work completely replaces OpenClaw?
显然，它们是独立的。
Obviously, they're independent.
是的。
So Yeah.
我并没有深入参与，所以不好评论这个。
I mean, I'm I'm not close to it, so I can't speak to Yeah.
OpenClaw 的路线图，但我觉得不会。
The OpenClaw roadmap, but I I don't think so.
我觉得，像 OpenClaw 这样由这个团队打造的优秀开源工具，始终会有它的需求。
I think that there's gonna be you know, there's always a need for, like, this, like, incredible, like, open source technology that that that team has built.
而且我觉得，我们在产品上可以从中获得启发，很多人其实更熟悉、也更常使用 ChatGPT，而不是 OpenClaw。
And I think that we can draw inspiration in the product and, you know, ChatGPT I think many more people have, like, heard about and used ChatGPT than have you have used OpenClaw.
如果我们能把 OpenClaw 的精髓带到 ChatGPT 上，那就算成功了。
If we can take the magic from OpenClaw and bring it to them, I think that'll be a success.
在 ChatGPT Work 这边，我们特别看重的一点是，用户一进来，就会和这个智能体展开对话，开启一个会话——不管你怎么称呼它。
I think that, like, one thing on the ChatGPT Work side that we feel strongly about is that, like, the core experience is that you come to this product, and you have a conversation, start a session, whatever you wanna call it, with this agent.
这个产品的神奇之处在于，你随时都能做任何事。
And the magic of the product is that you can do anything in that moment.
我们希望打造一个产品，让你不用点按钮，也不用跳转到别的地方。
And we would like to create a product where you don't have to click a button or to go to a different place, whatever.
你可以直接在这个地方，调用你财务软件或者其他任何应用里的功能。
And you can get whatever functionality exists in, you know, your your finances app or or any other product, like, in this in this one place.
这就是我们的目标。
And so that's the goal.
我们想要一个可扩展的系统，支持插件，让你能接入所需工具，完成财务任务；如果你在做科研，也能扩展系统，写代码、跑任务，都能流畅运行。
It's like, it we want an extensible system with plug ins where you can connect to the tools that you need in order to be able to accomplish, like, a financial task where you can, you know if if you're doing, like, science work, like, we have an ability to, like, extend the system and such that you can, like, write the tech and it and performs well.
当然，总有些专业工具在特定领域做得更好，但我们希望把尽可能多的神奇体验留在核心流程里。
There'll always be, like, products that we support that are best in class at those things, but we want as much of the magic as possible in that core experience.
是的。
Yeah.
你觉得在 ChatGPT Finance 里，你能完成以前在 Wealthfront 上做的所有事情吗？
Do you think that you can do everything you used to do with Wealthfront in ChatGPT Finance?
我其实试过。
I actually tried it.
我的意思是，ChatGPT 目前还不能替我保管现金和资产。
I mean, like, ChatGPT doesn't yet custody cash and and assets for me.
所以这一块，还没办法，现在还不行。
So so that part, no, not yet.
但当时我在那里的时候，我们确实研究过退休规划、财务规划、预算管理这些功能。
But I I mean, there was, like, a whole component of, like, retirement planning and sort of, like, financial planning and budgeting and stuff that that we were looking into when I was there.
现在有了财务插件，这些功能其实都能实现了。
And, like, with the finances plug in, like, that's all possible in terms of today.
至少在这些方面，我已经不需要原来那个工具了。
So I feel like at least that component's replaced for me.
我还没怎么用过这个插件。
I haven't really plugged it in yet.
我有点不敢去点开看结果。
I'm somewhat scared to look at the answer.
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。