OpenAI Codex 负责人谈产品工作的新形态 | Andrew Ambrosino

我觉得我们描述团队协作方式的一个关键点是，现在各个角色之间的重叠比以前多太多了。

And I think that one of the ways that we describe how the groups work together is that there's significantly more overlap in the roles than they used to be.

每个人不再被明确的界限定义——比如设计到哪里结束、工程从哪里开始。

And everybody's sort of defined less by the fence and the boundaries of where design stops and engineering starts.

而是更多地看他们实际工作的交集，对吧？

But more the average of where they're working, Right?

比如说，如果你把我们设计团队每个人做的所有事情平均一下。

So like, you know, if you average up all of the things that somebody on our design team does.

里面有很多写代码的内容。

There's plenty of code writing things.

也有很多属于产品方面的工作。

There's plenty of things that are product work.

但总体来看，他们的工作点其实都集中在中间这片区域。

But on average, there are dots over here.

对吧？

Right?

你在图上画出来。

You draw out on a diagram.

这也反映了整个流程。

And this sort of speaks to the process too.

尤其是因为Codex应用的整个架构都是通过自用循环形成的。

Especially because the entirety of the Codex app has been formed by the dogfooding loop.

我们所有人都希望尽可能多地在应用里完成工作，哪怕它还不是最好的工具，这样才能让它变得更好。

There is a desire among all of us to try to do as much as possible in the app, even when it's not the best tool, so that it can become the best tool.

所以很多设计工作，我们都是通过实际使用应用来做的，然后问：这个环节哪里出了问题？

And so a lot of design, we all work on by using the app and say, okay, what's broken about this?

我们经常不改进流程，而是通过优化产品来让流程本身变得更好。

This is a whole thing we do, which is that we we often don't improve our process, so that we can make the product better to do it.

这确实让人挺不自在的。

Which is a deeply uncomfortable place to be in.

但每周都在发生变化。

But, know, week to week it's changing.

我特别认同这一点，那你到底算什么角色呢？

I love this point so much that, like, what are you?

你的角色其实就是你花时间最多的事情的总和。

It's your role is the average of what you spend your time on.

如果你大部分时间都在做产品管理的工作，那现阶段你就是个产品经理。

If most of your work is PM y work, then okay, you're a PM for now.

对。

Yeah.

如果你主要在搞工程，那现阶段你就是个工程师。

If it's engineering, you're an engineer for now.

我觉得，OpenAI 是不是第一个把员工叫做‘技术成员’的公司？

I feel like, was OpenAI the first company to call people member of technical staff?

不是。

No.

我觉得这可能是施乐公司最早开始这么叫的。

I I believe that this this might have started with Xerox.

我第一份实习的公司叫Upthere，他们也是这么做的。

First company I interned at, it was company called Upthere, did the same thing.

这种做法早就有了。

This has been around.

不过现在更普遍了，但在以研究为导向的公司里，这算是一种传统。

But it's it's it's much more common now, but, you know, it is kind of a tradition in research focused companies.

明白了。

Okay.

懂了。

Got it.

所以这是从研究领域发展出来的。

So it emerged from research.

但我总觉得，这其实是个明显的信号，说明未来可能会朝这个方向走。

But I feel like it's such a, I don't know, sign of where things might be headed.

就是把所有人都称为技术成员，你的具体职能并不固定。

This idea of we're just gonna call everyone member of technical staff, your function isn't set.

你并不属于产品经理团队、工程团队或者设计团队中的任何一个固定类别。

You're not like in this bucket of the PM org or the engine org or the design org.

你觉得长远来看，我们都会朝这个方向发展吗？

Do you feel like that's where we all head long term?

你觉得这些职能还会继续存在吗？

Do you feel like functions will continue to exist?

毕竟，产品经理的能力、工程师的能力、设计师的能力还是存在的。

Like, there's still the PM skill set and the engine skill set and the design skill set.

是的。

Yeah.

而且大家还是会说，我是个设计师。

And people are, I'm a designer.

还是说，这种趋势其实就像人们常说的‘构建者’那样？

Or do you think this is like Like, people call it builder.

你觉得未来会迎来一场彻底的融合，每个人什么都做，这就是最终的方向吗？

Do you feel like there's this collapse coming where everyone's everything and that's just the future?

还是你觉得，我们将来还是会主要按受众来划分？

Or do you think we're gonna continue to be mostly divided up in the audience?

我有些担心的事情。

There are some things that I'm afraid of.

有些公司特别喜欢盲目跟风，别人说什么趋势，他们就立刻扑上去。

And I think that some companies like to be very extreme about getting on to the bandwagon of whatever people say is going to happen.

我觉得，取消角色这个概念的危险在于，它可能让人们对专业领域和可掌握的最佳实践失去认知。

And I think part of the danger in eliminating the concept of roles is that it can dangerously eliminate the idea that things are specialties with knowable best practices.

对吧？

Right?

我听过不少公司说，我们要取消产品经理这个职位。

I've heard a lot of companies be like, we're getting rid of the product role.

我觉得这想法简直太糟糕了。

Which I think is, by the way, a terrible idea.

结果人人都变成所谓的‘构建者’。

And everybody's just gonna be like a builder.

结果呢，他们就不喜欢产品领域里那些成熟的体系、真正经过验证的最佳实践，那些试过、失败过、最终沉淀下来的过程。

And then what happens is they don't like this whole discipline of product that's been built up and has like real best practices, real things that have been tried and failed and like real processes.

就觉得，哎，我写了几行代码而已，这些都不重要了。

Like that just gets abandoned because people are like, oh I wrote some code.

对吧？

Right?

这种状态其实挺危险的。

Like that's not a great place to be in.

我觉得，像‘这事儿不归你管’这种界限感，

I think that the boundary of like so and so, like this isn't your lane.

我是挺欢迎它消失的。

I I welcome that part going away.

但这里头有个平衡：不是每个人都能涉足所有领域，这是其一。

But there's a balance here where it's like not everyone can work on everything for one.

无论是广度还是深度，都不行。

Both in terms of breadth and depth rate.

这就是为什么管理者不会消失的原因。

Like this is why managers are not gonna go away.

不是每个人都能做所有事情。

Not everybody can work on everything.

而且每个领域都有其专业技能，很多工程师都没意识到，工程本身也是一种需要磨练的技能。

And also like every discipline has a skill component to it, which I think a lot of engineers are guilty of not recognizing that like, well engineering is a skill to it.

以为写代码很厉害，其他岗位就只是随便聊聊，这想法不对。

It's great in code and like other roles are just people vibing.

不是这样的，你得明白。

It's like, no, that's not how it works, right?

没错，你可以用Excel。

Like, yes, you can use Excel.

但光会用Excel，可不代表你能胜任财务团队的工作。

You But you cannot work on the finance team.

对吧？

Right?

就是这种事。

That is like, that kind of stuff.

对吧？

Right?

是的。

Yeah.

我觉得还得想想，你真想做这份工作吗？

I think there's also just like, do you want to be doing this work?

对。

Yeah.

我是不是想

Do I wanna be

做这个？

doing this?

现在更关键的是这一点。

I think more of it's that actually now.

对。

Mhmm.

换角色其实更容易。

It's like, it's easier to switch roles.

学最佳实践也更容易。

It's easier to learn the best practices.

没必要把你的工作表现和是否会用某个具体工具绑在一起。

It's easier to not tie your effectiveness in a role with the ability to use the exact tool.

对吧？

Right?

关键是能不能进入那种状态，搞清楚哪些方法有效、哪些没用，然后专心去做。

It's more of like, can you get yourself into this mindset, learn which things work and which don't, and then like focus on it.

对吧？

Right?

我以前花了很长时间觉得我不该当软件工程师，因为我对汇编语言或者背TypeScript语法根本没兴趣。

Like, I spent so long feeling like I should not be a software engineer because I didn't care about like assembly language or memorizing TypeScript syntax.

你知道吗，这些岗位一直都有些门槛，好像只要擅长某个工具，才算真正胜任这个角色。

You know, and it's like, there have always been parts of these roles that are that sort of gatekeeping that are like, well, no, this like being good at this role is being good at this tool.

我觉得这种观念正在慢慢被打破。

And I think that's what's kind of starting to erode.

我只是觉得，人们往往把这事夸大其词了。

I just don't I I think people take this, they hyperbolize all of this.

你们Codex团队现在是什么配置？

What does your team look like on on the Codex team?

有多少工程师、设计师和产品经理？

How many engineers, designers, PMs?

目前团队的人员构成大概是怎样的？

What's kinda like the makeup of the team right now?

每次有人问我Codex团队有多少人，你还记得我怎么回答的吗？

Every time people ask me like how many people are on the Codex team, do you remember my answer to this?

对。

Uh-huh.

我说吧，人数大概在十到几千之间吧。

I'm like, it's somewhere between 10 and a few thousand.

这答案听起来像是随便说的，但其实挺真实的，因为我们确实把这看作是所有人在这里工作的成果汇总。

I mean, it's like a fake answer, but it's it's real in that we do see this as the culmination of what everybody works on here.

所有涉及模型研究的工作都算在内。

Like everything that goes into model research.

所有关于模型如何擅长代码和浏览器操作的部分也都算。

Everything that goes into how you know, models are good at CUA and browser use.

包括模型性格的方方面面。

Everything about how you know, model personality.

还有前端基础设施相关的所有产品工作。

All of the product work around you know, front end infrastructure.

所有用户相关的内容，通通都是这个产品的一部分。

All of the user, like all of it is this product.

但与此同时，我们并没有每天接收成千上万人提交的、他们想做的任何功能请求。

At the same time, we are not accepting PRs daily from thousands and thousands of people on whatever they want.

我们团队有十几位工程师，设计方面大概有一半左右。

So we got a team double digits of engineers, probably half that on the design side.

产品方面人不多，不过产品这块儿更像他自己的一个独立阵地。

You know, a few product people, although you know, product tier is kind of more of his own defense play.

在Codex或者桌面端的团队里，大家普遍都有很强的自主性和审美判断，对吧？

And I think one thing that is very common among everybody on the Codex side or on the desktop side is agency and taste, right?

很多都是前创始人，或者在大公司里做类似创始人事情的人。

A lot of former founders or people who were at larger companies doing founder shaped things.

很多人眼光特别高，审美非常到位。

A lot of people with with immense taste.

在OpenAI，我们允许团队发展得相当大。

At OpenAI, we let teams get very large.

我们并没有说‘不能有管理层’。

So we haven't said, hey, there's no management.

但确实，团队规模都挺大的。

But like, the teams are quite large.

对吧？

Right?

主要是个体贡献者。

It's mostly ICs.

我觉得这样挺好。

And I think that's good.

你用‘区域防守’来形容产品工作，这很有意思。

You use this term zone defense for product work, and that's really interesting.

这其实也对应了设计领域的转变，你更像是在管理和协调。

It kind of maps to the design kind of shift also, just like you're there to kind of manage and coordinate.

能再详细说说这具体是怎样的吗？

Talk a little bit more about what that looks like.

对产品人员来说，‘区域防守’是什么样子的？

What does zone defense look like for a product person?

没错。

Yeah.

我和亚历山大聊过很多次这个比喻，意思是如果两个产品经理靠得太近，往往不是好现象。

And I have had a lot of conversations with Alexander about this analogy, which is that like, if two product people are working too closely, that's often not a good signal.

作为产品团队，你其实更希望有一种像斥力一样的分布方式。

And that like you kind of want, like as a product org, you sort of want to do this like force directed activity.

关键是要看哪里还缺人，哪里有空白。

Where you're like, where are the gaps?

尤其是在这个新环境下，内容筛选和方向对齐变得特别重要，因为大家到处抛想法，乱成一团，对吧？

Especially in this new world where curation and like, you know, steering alignment is a lot of things where you're like, there's a ton of chaos happening on people throwing ideas all over the place, right?

那种自上而下、规划一整年的老办法，现在根本行不通了。

The whole like top down, you know, year long planning thing not gonna work.

所以我们现在需要有品位的人，从想法萌芽开始就引导产品走向正确的方向。

And so now it's like, we need the taste makers to guide things from inception to what the product should be.

这就意味着你得做到全公司范围的覆盖。

And that means you basically want company coverage.

所以你要分散开，想想谁最适合做什么。

And so you spread out and you say, alright, who's best at what?

咱们得保持一点距离，这样才能全面覆盖，对吧？

Let's create some space between us so that we got full coverage, right?

事情本来就是这样的。

And that's kinda how it goes.

然后你再填补空白，比如我们会说：咱们要招那些有产品思维的工程师。

And then you fill in the gaps and you're like, look, we wanna hire engineers who are product minded.

我们不希望出现一堆人埋头写代码，最后还得整个团队来审查代码是否符合产品逻辑。

We don't want it to be like, you know, we've got a bunch of people writing a bunch of code that needs full team reviewing it for product coherence.

我们希望每个人都具备这些能力。

We want everyone to have these skills.

但我觉得大家需要深入专精的方向，得变了。

But I think what people go deep on has to change.

对吧？

Right?

我跟像你这样的人聊了这么多，越来越明显的一个趋势是：现在最有价值的人，就是那种能把一个想法从零做到落地、还懂判断什么是好东西的人。

This This is definitely a thread I've been noticing over and over with talking to folks like you is the the most valuable person right now, one of the most valuable, someone that could take an idea from idea to done with the taste to know this is great.

是的。

Yeah.

就像在整个过程中一直引导着，把每件事做到极致。

Just like shepherding throughout this obsession with making it awesome.

就是你刚才说的那种高主动性、高品位的人。

Like this kind of high agency, high taste person exactly as you described.

这是否就是你心目中，我们该招什么样的人，谁会在这个新世界里表现得特别出色？

Is that is that kind of the way you think about here's who we're hiring, here's who's gonna do really well in this new world?

对。

Yeah.

我觉得这现在才是核心所在。

I think that that's the core piece right now.

这也反映了我对个人贡献者和管理角色的看法——管理并不会消失。

And it also speaks to how I sort of see IC versus management, which is that it's not that management is going away.

也不是说每个人都得当个人贡献者，但如今每个人某种程度上都得兼而有之。

It's not that everyone's an IC, but like everyone's kind of both now.

对吧？

Right?

如果你是技术人员，可不是一个字符一个字符地敲代码吧？

If you're an IC, you're not typing code out character by character, right?

你其实是在管理一些东西。

Like you are managing something.

你在管理各种代理，管理那些正在发生的工作，对吧？

You're managing agents, you're managing you know, like you're managing work that is happening, right?

这些工作最终汇聚成某个成果。

That comes together to do a certain thing.

如果你是团队管理者，其实也在做同样的事，只是粒度不一样而已，对吧？

If you're a manager of teams, you're doing the same thing, just at a different, like different granularity, right?

我通常会找那种，当然得精通专业领域，但更重要的是有判断力，知道就算给你无限资源，也不能随便应付了事。

I generally look for like, obviously command over the discipline, but then the taste to say like, hey, you're gonna have unlimited tokens and I don't, like we can't just be doing slop.

你得能在信息爆炸的世界里，分得清什么是真正有价值的信号，什么是噪音。

Like you need to be able to determine what signal, what's noise, like, in a world of just infinite content.

提到规划。

Mention planning.

以现在的变化速度，制定路线图变得非常困难。

At the pace things are moving, it's become very hard to plan road maps.

是的。

Yeah.

我猜在你的领域里尤其如此。

I imagine especially in your world.

在这方面，人们总是对我特别不满。

People are very frustrated with me all the time on this.

没错。

Yes.

因为东西一直在上线，情况一直在变。

Because things are just constantly shipping, things are changing.

对吧？

Right?

你们团队是怎么做规划的？

How do you plan on on on your team?

你们一般会往前想多远？

What's kinda like how far ahead are you thinking?

一个计划具体是什么样子的？

And what does a plan look like?

是用表格来做的吗？

Is it like a spreadsheet?

是用 Markdown 文件吗？

Is it MD file?

那计划的最终产出到底是什么？

What what's kind of the output of a plan?

是的。

Yeah.

我觉得我们在这一块并没有什么特别创新的做法。

I don't I don't think we do anything revolutionary on that.

我们在规划上并不玩什么花活。

We're not clever about planning.

我觉得基本道理是，时间越近的事情，就越需要具体细节。

I think like the basic gist is, the shorter term something is, the more detail it needs.

我们不是不规划九个月后的事，只是那时候的内容必须保持模糊。

And then it's not that we don't plan for nine months out, it's that that just has to stay very hazy.

因为现在给九个月后的计划加太多细节，那都是虚假的精确。

Because any amount of precision that you add to a nine month plan right now is false precision.

纯粹是在浪费时间。

And like, you're just gonna waste time.

但你当然可以说点什么，对吧？

But you can say stuff, right?

但那些都不是我们真正规划好的内容。

But like, nothing that we planned.

我觉得研究这块不一样，我这里说的不包括研究。

I think research is different, so I'm not speaking for research here.

但在实际应用这边，像十一月能规划好的东西，可能十二月看着还成立，但实际根本不是那么回事，对吧？

But like on the applied side when we do products like anything that you could have planned in November may have been true for December, but like isn't what happened, right?

所以这事儿挺难的。

So it's hard.

真的很难做规划。

Like it is really hard to do planning.

我们通常得搞清楚，模型在什么时间点能实现哪些能力。

We generally need to know like what do we think models are able to do on what timeline.

我上一家公司就经历过这种转变：我们开始用模型来驱动功能，结果产品流程就崩了。

And my last company, I kinda saw this shift where we were starting to use the models to drive features and the product process fell down.

最后我们只能这么做：先把未来一两年内所有可能想做的东西列出来。

It basically had to be like, let's list out all of the things that we think we are interested in doing for the next year or two.

然后全都做个原型，看看哪些现在就能用，剩下的就先放着，慢慢沉淀。

Let's prototype all of them, decide which things are ready now, and then just let the others sit and bake.

每次模型有重大突破，就拿这些搁置的点再试一遍，换上新的模型。

And then every time there's like a new leap in models, let's try that thing again with it swapped out.

因为判断功能好坏的依据，完全是看它够不够智能，而不是它的形态如何。

Because like the whole premise of whether features were good or not were based on whether they were smart enough, not the shape of them.

这让我想到一个关于Codex应用的绝佳例子。

So this is a great story about the Codex app.

我非常确信，如果我们二月发布的这个Codex应用，早在十一月就上线了，它在市场上肯定会彻底失败。

I like I am very confident that the Codex app that we released in February, if that had been ready in November, it would have absolutely failed in the market.

而十一月到二月之间，唯一的不同就是模型进步了，对吧？

And the the only difference was the models between November and February, right?

我觉得这里面大有文章。

And I I think like there's a lot to that.

同一个产品，形态一点没变，但只因为推迟了几个月上线，结果就天差地别。

That this product with the exact same shape, I think would have like, it's it's outcomes were totally different depending on just a few months of timing.

本节目由Mercury赞助播出。

This episode is brought to you by Mercury.

一种截然不同的银行服务，已受到超过三十万创业者青睐，现在更推出了命令行功能。

Radically different banking loved by over 300,000 entrepreneurs and now with command.

我已经是Mercury的客户六年多了。

I've been a customer of Mercury's for over six years.

我从来没想过要换银行。

I have never once thought about leaving.

Mercury就是当银行由产品人而不是银行家来打造时会变成的样子。

Mercury is basically what happens when banking is built by product people, not by bankers.

他们让开票、转账、给团队成员设置虚拟卡片变得特别简单，甚至可以说很有趣。

They make it so easy, dare I say fun, to send invoices, move money around, set up virtual cards for folks on my team.

你的银行有API、原生命令行工具，或者支持AI的MCP服务器吗？

Does your bank have an API, a terminal native CLI, or an AI ready MCP server?

我觉得没有。

I don't think so.

最近，他们推出了Command，一个直接内嵌在Mercury里的对话式界面，相当于你的财务助手。

And just recently, they launched Command, a conversational interface built directly into Mercury, which acts as your financial operator.

我一直在用Command来转账、分析我花得最多的类别、查看现金流，就在今天，我还用它查了过去一年从某个赞助商那里赚了多少钱。

I've been using command to transfer money around to figure out what categories I've been spending the most money in, analyze my cash flows, and just today, I used it to find out how much I've made from a specific sponsor over the past year.

我只要问一句：过去一年我从X那里赚了多少钱？

I just ask, how much have I made from X over the past year?

十秒钟后，答案就出来了。

Ten seconds later, have an answer.

这简直太酷了。

It is so freaking cool.

访问 mercury.com 了解更多，几分钟内在线申请即可。

Visit mercury.com to learn more and apply online in minutes.

Mercury 是一家金融科技公司，银行服务由 Choice Financial Group 和 Column NA 提供，两家均为 FDIC 成员。

Mercury is a fintech company not an FDIC insured bank, banking services provided through Choice Financial Group and column NA, members FDIC.

这个播客里确实有一条主线，就是先做出还没完全成型的东西，等模型变强了，它自然就行了。

This is definitely a thread on this podcast is build things that are not yet working, then will work when the model gets better.

还有另一条主线，是关于野心的。

And there's this kind of other thread of ambition.

对你要接手的事情，要有更大的抱负。

Be more ambitious with the things you take on.

所以，这是一种你处理事情的方式吗？

So is this just like a way you approach things?

就是先动手搞一堆可能还不太行的东西。

Is just like, let's just build a bunch of things that may not work yet.

先放着，等技术跟上再说。

We'll just have them around and wait for a model to catch up.

你是这个意思吗？

Is that kind of the approach?

对。

Yeah.

我们确实有不少这样的情况。

Think we have a lot of that.

我觉得难点在于，你得清楚现在到底处在设计的哪个阶段。

I think sometimes the challenge is like, you have to be very clear again about what stage of the design process that's in.

人们还总带着一种惯性思维：我写完代码了，那就该上线了。

People still have this muscle memory of like, oh, I wrote the code for this thing, therefore we should put it out there.

不是，不是，不是这样的。

It's like, no, no, no.

这意味着我们现在有了一个可以用来测试未来模型的成果，对吧？

That means you have an artifact now that we can test against for future models, right?

就像我们应用里的内置浏览器，这种情况也发生过，对吧？

This happens with the in app browser, in the app that we have, right?

我们之前确实有个能用的版本。

Like, we had a kind of a working version.

再说Atlas，我们那时候已经在Atlas里运行了智能代理，那确实挺厉害的。

I mean, go back to Atlas, we had agent working inside of Atlas and you know, that was pretty cool.

在ChatGPT里，我们之前还有个Operator功能，对吧？

We had operator before that in ChatGPT, right?

那个没成功。

That didn't work out.

这个想法其实挺棒的，Operator、Atlas、Codex和ChatGPT之间其实有一条清晰的线索可以串联起来。

Very cool idea like there's some thread that you can draw between operator, Atlas, Codex, ChatGPT.

这本质上是同一个功能，但用不同的智能重新发布后，结果就完全不一样了。

That's like fundamentally the same feature, but the re releasing of it with different intelligence totally changes the outcome here.

所以我总是劝大家别固执地觉得‘这不行，所以这功能不好’。

And so I push people not to be stubborn about like, no, this isn't working so it's a bad feature.

其实可能只是还没到时候。

Like no, it might not be ready yet.

尤其是在研究领域，大家总想追求最宏大的目标，觉得‘极限情况下，模型本来就应该能做到这一点’。

There's also this aspect of, especially in research, there's always a desire to be the most ambitious and to say, okay, but at the limit, the model can just do this.

但在产品层面，这根本行不通。

And that just doesn't work on the product side.

如果你回看Codex最初发布时的样子，其实它就是Codex Web，根本没法好好交互。

Like if you if you go back to the original Codex release, basically what it was, is it said, it was Codex Web and it wasn't good for interacting with.

你给模型一个任务，它就自己去跑，跑完了再把结果丢给你。

It was like, you give the model a task and it's gonna go off, do the task, come back to you with it finished.

听起来也没多惊人。

Like, doesn't sound that radical.

问题是，它完成任务的效果并不好。

The problem is like it didn't do the task that well.

它确实能写代码，质量也不错，但这种使用形式太早了。

Like it wrote code, it was it was it was good but it was like that form factor was too early.

后来推出的云代码却是完全本地化的，根本没连上云端。

And then the cloud code comes out totally local, like not hooked up to the cloud.

它不假装自己有多智能，对吧？

Doesn't pretend to be as it's not as AGI-pilled, right?

它会主动问你问题，然后安静地等着你回应。

It's like gonna ask you questions, it's gonna sit there.

你不能把生活全丢给它去处理。

You can't just delegate your life to it.

那样反而好很多，对吧？

That worked way better, right?

因为这才是重点——模型本来就应该这样发挥作用，对吧？

Because that's the point that the models were out there, right?

所以我们当时太沉迷于通用人工智能了，时机还没到。

So we were like, we were too AGI pilled for the moment.

我经常回想起这个教训，尤其是在做这类事情的时候。

And I think like, I think about that lesson a lot on this stuff.

以前呢，市场失败会告诉你产品该长什么样、该怎么传达它的价值。

It used to be that you know, failing market told you all these things about the shape of the product, about the communication of the product.

但现在不一样了，你可能得发布六次，这东西才能真正跑通。

And now it's like, no, you might need to release this thing six different times before it works.

而且产品的形态可能根本不会变。

And that might, like the shape might not change at all.

这事儿真的挺有意思的。

There's like, it's so interesting

听到你现在做产品要考虑这么多变量，真是长见识了。

to hear about all the variables you have to think about building product now.

得考虑模型的发展节奏、研究进展，还有它变得有多聪明。

There's the timeline for the models and the research and how smart it gets.

现在连很多人还无法理解，原来软件还能这样在云上开发，这才是未来。

There's like people's ability to even understand this is how you could build software in the cloud and this is the future.

得让大家提前适应这个新未来，然后才能看清团队到底能做出什么。

Like, get people prepared for this new future and then just what you can build as a team.

我特别喜欢Codex这个例子，因为它又回到了野心这个话题。我想听听你有没有类似的感受——不妨更大胆一点，因为这些模型能做的事，远超你想象。

And I love that Codex example because it comes back to this idea of ambition, and I wanna hear if there's anything there for you of just this thread of just be more ambitious because these models can do so much more than you can imagine.

有时候目标太超前，市场还没准备好，但你至少该想想，怎么激励团队更有野心，毕竟现在做那些过去觉得难如登天的事，反而变得容易多了。

And sometimes it's too ambitious for the market and they're not ready for it, but you think about that at all just like pushing your team to be more ambitious because it's so much easier to just do things that maybe felt crazy hard in the past.

是的。

Yeah.

这正是核心难题。

This is a core challenge.

一旦产品或功能上线了，人们就很容易盯着小问题，不断微调。

Once once there's a product that exists or a feature that exists, it's really easy for people to find paper cuts and Micro optimize.

他们确实应该这么做。

And they should.

而且人们在推特上总会提醒我们这一点，我很感谢他们这么做。

And people on Twitter like to remind us of this and I thank them for that.

大家应该专注于现有功能，让它们更稳定、更出色。

People should be focused on the features that exist and making them more reliable and better.

但这也正是我们这里鼓励自下而上探索的原因。

But this is why we also have a culture of bottoms up exploration here.

因为有时候，就像Codex应用当初某种程度上颠覆了ChatGPT那样。

Because sometimes, in the same way the Codex app came and disrupted ChatGPT in some way, right?

未来也一定会有什么新东西来颠覆现在的成果。

This thing will get disrupted by future effort.

这本身就是设计的一部分——一个团队很难同时兼顾颠覆性创新和产品稳定性这两件事。

And that's that's part of the design is that like, you can't always as one team be good at both the disruptive piece and the like maintaining a product and its quality piece.

但总归可以设计出一套流程，让两者都能兼顾。

Some point, you could design a process that allows for both.

稍微宏观一点来看。

Kinda zooming out a little bit.

想想看AI在产品开发方式上带来的演变，简直不可思议。就像你说的，我们过去全是手工写代码，一点一点精心打造，现在却变成AI完全代写，甚至你刚才说，现在编程其实是引导AI。

If you think about the progression that we've been on of AI impacting how we build product, It's, like, insane how far we've come from, as you said, we used to write all our code by hand, like, artisanally created human code, to AI writing a 100% of our code, to you actually put it this way that now, like, coding is steering the AI.

你想想看，我写的代码里有多少是AI生成的？其实关键不在于写了多少，而在于我得多少次把AI往正确的方向带——这才是新时代的编程。

And, like, when you think about what percentage of my code is written by AI, it's almost like how many times did I have to steer it in the right direction is the AU version of coding.

现在有了智能体、循环这些新东西，据你观察，目前大家在开发上最前沿的做法是什么？

And now that there's, like, agents and loops and all these things, what's kind of the latest frontier from what you've seen of how people are building?

循环？

It loops?

还有没有其他更前沿的做法？那些最激进的AI驱动团队，他们现在到底是怎么运作的？很多人可能还不知道。

Is there something else of just like the most AI pill AI forward teams, here's how they operate now that people may not be aware of?

是的。

Yeah.

循环都过时了，老兄。

I mean loops are so last week, man.

我们之前就聊过这个。

I mean we we talked about this.

你想想，大家一直都在问一个问题：到底有多少产品代码是AI写的？

You know, one of the big questions is always, well how much of the product is AI written?

这个问题其实很难回答，因为如果你还用去年的标准来看，那现在我们整个产品代码都是AI写的。

And it's always hard to answer that question because if you're using the goalposts from last year, it's like well 100% of our product right now is AI written code.

所以真正该问的是：好吧，那又怎么样？

So the question is more like, well, okay fine.

代码是有人监督着写的，还是完全自主生成的？这才是关键。

Is the code written supervised versus unsupervised, right?

这完全是两码事。

And that's like a totally different thing.

我很欢迎这种标准的变动，因为这说明我们在真正推进产品。

I welcome the moving of goalposts because that means we're making product progress.

这里头已经有不少人尝试过完全自主开发的软件了。

There have been a lot of explorations here around like autonomously developed software.

各种自动化工程的探索，五花八门，我就在想，要是你一觉醒来，系统自动把代码库清理一遍，清理掉那些垃圾代码，会怎么样？

A lot of like harness engineering stuff, a lot of different explorations, I'm like, okay, well what if you came in overnight and did garbage collection of the code base to clean it up, right?

我觉得现在所有模型都面临一个问题，就是它们往往会增加系统的复杂性。

One thing that I think all models suffer with right now is just they they usually increase complexity.

如果哪家公司还在做研究，拜托让模型更擅长删除代码吧。

If research is listening at any company, please make the models better at deleting code.

但当你想把开发完全交给自动系统时，这个问题就凸显出来了。

But you know, that becomes a problem right now when you try to put development completely on autopilot.

这既涉及人的方面，也涉及代码库本身。

And it's both on the the human side and the code base side.

比如说功能需求，对吧？

So like feature requests, right?

你怎么教模型分辨哪些功能该做，哪些该忽略，哪些该合并起来重新设计一下？

You know, how do you teach a model which features to build, which ones to ignore, which ones to kind of like group together and reframe a little bit.

你怎么教模型学会构建正确的抽象结构？

How do you teach a model, how to build the right abstractions, Right?

所有这些都在慢慢变好。

Like all of this is getting better.

我觉得现在还没到那一步，还没办法直接设个循环，让系统自己去改进应用，监听推特、Slack 和邮件。

I don't think we're at a place yet where we're like, we're just gonna set up a loop that's like improve the app, you know, and listens to Twitter and listens to Slack and listens to email.

我们离那个境界还远着呢。

And like, we're not there yet.

但我们正在努力朝这个方向推进。

But we are we are we're trying to make it happen.

你觉得我们能实现吗？

Do you think we'll get there?

你觉得有一天会达到那种状态吗？系统自己就能不断成长、胜出？

You think we'll get to a place where it's just like, grow, like win?

或者说目标就是赚钱，让我赚它一个十亿美元。

Slash goal, make money, like, make me a billion dollars.

对。

Yeah.

在市场中胜出。

Win win the market.

我也不确定，老兄。

I don't know, man.

嗯，我是说，有可能吧。

Like, I Yeah.

我从不轻易说‘是’或‘肯定’这种话。

I'm not in the business of saying Yeah.

从不，或者总是，这类说法我都避免。

Never or always or whatever.

对。

Yeah.

作为产品负责人和工程负责人，你是怎么在工作中使用人工智能的？

How are you using AI in your work as product leader, engine leader?

对。

Yeah.

你平时用AI的方式里，有哪些是别人可能根本没想到还能这么用的？

What are some ways that you use it that maybe people may not be aware they can use the app for.

是的。

Yeah.

我觉得我现在拥有的是世界上最好的工作。

I think I have the best job in in the world right now.

但其中特别有趣的一点是，当初我们开发原始的 Codex 应用时，我个人的目标就是做出一个我自己写代码时会用的工具。

But one of the things that makes it very fun is that when we were developing the original Codex app, the goal for me personally was to make it the thing that I wrote the code with.

对吧？

Right?

我当时就想，我得把这个工具做得足够强大，让我能用它来开发 Codex 本身。

I was like, I need to make this so good at development that I can build the Codex app with this.

那时候的 Codex 应用本身就是一个开发工具。

And the Codex app at that time was a development tool.

对吧？

Right?

我们很快就进入了内部自用的迭代循环，因为你总会有那种个人体验的时刻——比如，哎，这事我怎么就做不了呢？

And we did that like super quick dogfooding loop because you've got your personal dogfooding loop where you're like, oh, like I can't do this thing.

我得把这个修好，这样才能搞定。

I should fix that so that I can do the thing.

现在我能做这件事了。

Now I can do the thing.

现在我能做更多事了，对吧？

Now I can do more things, right?

我们后来就把这个发布了。

You know, we released that.

接下来的挑战是，发现大家开始用它来做一些不一样的事情，对吧？

And then the next challenge was, hey, people are starting to do some different shaped things with this, right?

于是我得把这东西扩大，得招几个人来帮忙。

And now I, you know, need to grow this and so I need to hire a few people and help.

所以，就在应用功能需要转变的同时，我的角色也变了。

So then like my role changed at the same time that the role of the app needed to change.

我就想，好吧，我得在这儿多做点产品探索了。

So I'm like, okay, I need to do more product discovery here.

我得弄清楚该怎么跟踪每个人的工作进展，并及时纠正偏离方向的事情。

I need to figure out the right loops for seeing what everybody's working on and steering things that are off track.

所以突然间，我就开始用Codex这个应用来干这个了，对吧？

And so all of a sudden, that's what I started using the Codex app for, right?

我其实还是在写代码。

I did still write code.

我一直在调整自己的使用方式，让它更贴合我们想解决的问题。

It's like I've I've tried to align my own usage of it with the problem that we're trying to solve.

对吧？

Right?

现在我觉得，我得做个表格，把这套机制梳理清楚。

And now I'm like, I need to build a spreadsheet that models this out.

我得深入调研一下，看看这个领域之前都做了哪些努力，为下一个版本做准备。

I need to kind of do, you know, internal deep research on all of the efforts that have gone into this area of research for the next version of this.

五月份左右发布了一次或一系列更新，引入了应用内浏览器。

There's a release or a series of releases in May ish that introduced the in app browser.

把电脑使用和成果创建整合到Codex应用里。

Computer use and artifact creation to the Codex app.

那应该是我们为大多数人推出的Codex版本。

That was I think our Codex's for almost everyone release.

大家都懂‘氛围编码’这个词。

And everybody knows the term vibe coding.

我觉得那是我们第一次做氛围协同的发布。

I think that was like our first vibe coordinated release.

我当时有个Notion文档，列出了所有要做的事，然后我就开始自动化处理，从Pull Request和Slack频道里收集更新，更新状态追踪表——现在这已经很常见了。

We're like I had a notion doc somewhere with everything that needed to happen and I was like automating, like going out to gather updates from pull requests from Slack channels and like updating the status tracker and like now this is pretty common place.

但那时候，我感觉自己正站在产品发布管理的最前沿。

But at the time, felt like I was at the bleeding edge of like, how to manage a product release.

简单说，我用Codex的方式，其实就是看我的工作变成了什么样，然后怎么让这个工具能帮我搞定所有事。

In short, the way that I use the Codex app is basically like, what has my job grown into and how do I make this thing able to do everything I need to do?

每天早上起床，我会看我收到的每日简报，那是从我加入的三千多个Slack频道里汇总出来的，告诉我哪些事情需要我关注。

I will get up in the morning, I will see the daily brief that I have from like everything from the 3,000 Slack channels that I'm in, which things need my attention.

我可以回个消息，说：好吧，给我五个问题，我来回答，这事儿我能搞定。

I can kind of message back and be like, alright, give me five questions and I'll answer them and I can do that.

你是怎么设置的？

How do you set that up?

一般人要怎么搭建这个流程？

What's like the workflow for somebody to set that up?

明白了。

Yeah.

听起来太棒了。

That sounds amazing.

再说一遍，我们对很多这部分内容还在探索阶段。

Again, think we're still in the discovery phase on a lot of this.

所以现在，我正在做一个自动化程序，或者说是定时任务，让它去浏览我的 Slack 频道。

And so right now, it's like I'm making an automation that says, or scheduled task that says like, you go through my Slack channels.

这些是我关心的、最重要的内容。

These are the things that I care about and think are most important.

其实我还在慢慢明确这一点。

You know, so I'm still kind of defining that.

比如哪些事情值得关注，分门别类，附上背景信息，这些都会设成自动任务。

Like these are things to watch out for different categories like here's some context and you know, I'll I'll you know, that'll get set up as a automated task.

刚开始运行的几次，我可能得稍微引导一下。

And then first few few times it runs, I might need some steering.

幸运的是，这个应用让我不用费劲去改指令。

And luckily with this app, know, I don't have to find out how to edit the instructions.

我只要说：下次运行时，能不能优先处理这个？

I can just be like, hey, next time this runs like, can you please worry about this instead?

或者，能不能把这条工作流的优先级调低？

Or can you deemphasize this work stream?

嘿，刚才发生了一件事，但简报里没提到。

Hey, this thing happened and it didn't come up in a brief.

能不能确保这类事情以后都能被覆盖到？

Like, can you make sure that stuff in the shape?

所以我可以在过程中慢慢引导它。

So I can kind of coach it along the way.

它会自动调整提醒我的方式，诸如此类的事情。

It'll update the way that it notifies me, stuff like that.

太棒了。

Amazing.

我觉得将来，这一直是聊天机器人设计的一个核心问题。

I think in the future that like, this is this has been a core problem with the the chatbot shape.

对吧？

Right?

我知道该怎么设置它。

Is that I know how to set this up.

我有时间去设置，因为对我来说，这本身就是产品探索的过程。

I have time to set this up because for me, it's product discovery to set it up.

但如果你不是在OpenAI工作，也不是在开发这个系统，谁也不想花时间去搞懂这些复杂的东西。

But if you're not working at OpenAI, not developing this, like, you don't wanna have to figure out all this stuff.

我们真得好好想想该怎么设计这种模式。

Like, we we we need to like figure out that that shape of things.

对。

Yeah.

我听到的是，很多人没意识到你的应用其实可以像OpenClaw那样工作。

What I'm hearing here is, I don't think people realize that your app can act a lot like OpenClaw.

是的。

Yeah.

人们之前对这个特别兴奋，就是你只要跟它说一声，设置好之后，它每天帮你查看一下，然后告诉你进展如何。

You was people were so excited about, like, you just talked to it, set up this thing, check on this thing for me every day, and then tell me what's going on.

没错。

Exactly.

对。

Yeah.

这种功能正逐渐融入各种产品中，这真是太棒了。

Like, it's starting to become a part of all these all the products, which is amazing.

所以用户要设置这个功能时，只需要在应用里直接说：‘我想设置一个自动化，帮我查看Slack，完成这些任务。’

So the way somebody would set this up is they just talk within the app and say, wanna set up an automation to do this, look at my Slack and these are the things I I wanna do.

是的。

Yeah.

很好。

Great.

对。

Yeah.

应用会回应说：‘好的，明白了。’

And the app will say, you know Yeah.

它会帮你自动设置好。

It'll set it up for you.

如果还没接入Slack，它会问：‘要连接Slack吗？’

If it doesn't have a Slack connector, it'll say, it's gonna hit on the Slack connector, yes or no?

你只要回答‘是’就行。

You can hit yes to that.

是的。

Yeah.

至少我们可以做到，如果你不知道怎么在应用里操作，直接问它就行。

Like, the least that we can do is make it so that if you don't know how to do something in the app, they can just ask it.

是的。

Yeah.

对吧？

Right?

是的。

Yeah.

我觉得光这样还不够，但至少这是我们该做的底线。

I don't think that's enough, but I think that's the least we can do.

对。

Yeah.

举个例子，我做过一个小应用，能自动过滤收件箱里的垃圾邮件。

A good example is I built this little app that filters spammy email from inbox.

所以每收到一封邮件，我都是用Codex做的，它会自动分析这封邮件是不是那种我不想要的垃圾推销邮件，然后给它打标签，移到别的地方去。

So every email that comes in, and I built this in Codex, every email that comes in, it looks at it and decides is this unsolicited kind of cold email y stuff that I don't wanna look at and labels it and puts it somewhere else.

设置这个功能的时候，有个步骤得去Google Cloud控制台，配置一堆Hub、Sub、Hubbub的API和触发器。

And to set that up, one of the steps was you have to go into, like, the Google Cloud console and set up all these hub sub, hubbub API things and triggers.

不知道你有没有用过那个界面。

I don't know if you've ever used that interface.

那玩意儿真是烦得要命，没错。

It's, like, so annoying and so Right.

我当时就想，等等。

So I was like, wait.

要是我直接问你来帮我做呢？

What if I ask you to do it?

然后我就想，行啊。

And I was like, okay.

太好了。

Cool.

帮我做一下。

Do this for me.

你只需要描述一下怎么操作电脑就行。

And it just you describe computer use.

我以前从来没在电脑上见过这种事。

I've never actually seen this happen on my computer before.

它直接接管了我的电脑，自己开始运行了。

It just takes over my computer and starts going there.

管它有没有接口呢，哥们。

It's like, don't care if you don't have a connector, man.

我直接点就是了。

I'll just start clicking.

对。

Yeah.

它自己就搞定了。

And it figures it out.

看着它自己运作，真是有点不可思议。

It's crazy just to like watch it doing its thing.

设计连接器之间的决策边界挺有挑战的，得决定什么时候用内置浏览器，什么时候用已连接的Chrome扩展，什么时候直接用电脑操作。

Designing the decision boundary between connectors, when to use the in app browser versus your Chrome extension that's connected versus computer use.

是的。

Yeah.

挺有意思的。

It was interesting.

全靠慢慢摸索出来的。

And all done through just like feeling it out.

前几天我看到一个很棒的Twitter帖子，里面详细讲了这三种方式各自该怎么用。

I saw a great Twitter thread the other day where they describe all these three and what you use it for.

对。

Yeah.

这个人说得真到位。

So this person described it really well.

这些个人工作流程特别有意思，因为有些真的特别管用。

These personal workflows are really interesting because some of them really click.

比如很多人在尝试各种各样的方法。

Like some of the you know, people are trying all sorts of stuff.

每个人都在打造自己的专属系统。

Everybody's making these personal systems.

你问这儿每个人是怎么做的，答案全都不一样。

You ask everybody here what they do and everything's gonna be different.

但过一阵子，一些共同的模式就浮现出来了，我们会想，嘿，这个应该做成应用里的核心功能。

And then like certain themes arise and we're like, you know what, that should be a first class experience on the app.

既然大家都这么 setup，那我们干脆直接把它做好。

Like we should take this thing that everybody seems to be setting up and just make that work.

我觉得现在的记忆工具还处在早期阶段，很多人，包括其他公司的人，都在说：我搭了个 Obsidian 或 Notion，自己手动配置怎么构建思维宫殿。但说实话，你根本不该这么折腾。

And I think memory is sort of in the shape where we've had a lot of people and a lot of people at other companies too are like, well I set up an obsidian base or a notion area and I tell it how to basically build my mind palace and how to put It's like, I don't know if everyone Like you shouldn't have to do that.

应该有个内置的记忆功能，自动帮你搞定这些，对吧？

Like there should be a memory feature that does that for you, right?

这太笼统了。

That's pretty generic.

所以呢，这算是一种情况。

So that's you know, that's one.

但还有其他情况，比如你工作中的流程，有些东西确实值得你去设置一下。

But then there are other things like, they're your process of your job and there's something that like, yeah, you should set that up.

但我觉得我们一直在观察，到底哪些方法对个人有效？

But I think this is, we're constantly waiting through like, what's working for individual people?

哪些功能应该纳入产品，哪些只是个人的工作习惯？

Like what should enter the product versus say like, no, that's just how you do your job, right?

哪些该成为基础功能，哪些不该。

What should become a primitive, what shouldn't.

对。

Yeah.

没错。

Yeah.

这正是你之前提到的品味和判断力，就是靠这些来衡量的。

This is the the taste and judgment you spoke of earlier, know, citing these things.

我想聊聊浏览器使用这个点，因为我觉得很多人没意识到它有多强大，能派上什么用场。

I wanna talk about this browser use piece a little bit because I think people don't realize how powerful this is and what it could be used for.

说到这个，我不确定你有没有听过丹·希珀在播客里的那期。

Reminds me, I don't know if you watched when Dan Shipper was on the podcast.

他当时提到了Every这个工具。

He had this from Every.

他还预测说，我们以后会直接用Codex来运行我们的SaaS应用，就在对话界面里。

He had this prediction that we're gonna start using Codex to run our SaaS apps inside of Yeah.

而不是打开Chrome浏览器。

So instead of going in Chrome

我知道他每天都会在Slack上找我，让我帮他做这做那。

I know he slacks me about this every day, ask ask him for stuff.

你觉得未来真会变成这样吗？我们直接在Codex里工作，用Notion、Linear、Salesforce，全靠智能助手帮我们推进？

Do you feel like this is where things go, where we're just working within the Codex app using Notion and Linear and Salesforce inside with your agent kinda helping you along?

还是你觉得这是个完全不同的方向？

Or do you think that's a kind of a different direction?

确实如此。

It's been Yeah.

这事儿挺有意思的，毕竟我们之前已经试过好几种类似浏览器的形态了，比如Operator、ChatGPT的代理模式、Atlas，现在又在桌面应用里内置了浏览器。

It's been really interesting because obviously we've had a few tempted browser shaped activity, right, And operator and ChatGPT agent mode and Atlas and now we have the in app browser inside of desktop app.

我们还能安装Chrome扩展，让应用和Chrome连接起来。

We also have the ability to install a Chrome extension where the app connects to Chrome.

这方面我们尝试过很多种形式。

We've had a lot of shapes with this.

我们也从中积累了不少经验。

And I think we've learned a lot of different things.

这里面牵扯的东西很多。

There's a lot at play.

还有很多特别枯燥但很重要的因素在起作用。

There's a lot of really boring things at play.

我们最初推出这个应用的时候，是个 Electron 应用。

Like, you know, we originally launched the app, it's an electron app.

在应用里用内嵌浏览器能做的事，其实挺别扭的。

The things that you can do with in app browsers in there, it's like kinda janky.

当初内嵌浏览器只是为了开发和测试前端，我们还跟大家说，这玩意儿真不是给普通用户用的，就是个开发者工具，对吧？

So we have the in app browser was for development, it was for testing your front end on development and we were like, it's not really for anything else guys, like it's it's developer tool, right?

后来我们转到了 OWL 技术栈，这套东西原本是给 Atlas 浏览器用的。

And then we we switched over to our OWL Stack which had powered the Atlas browser.

现在它支持多标签了，还加了企业级安全功能，你只要登录一次，就能访问所有网站了。

And so now, know, it multi tab and we've got enterprise security so that you can actually log in to all your websites if you're, you know.

我们一直在不断优化这个功能。

So we've been iterating on this.

我一直觉得最难的是，这个浏览器到底该长成什么样？

I think the tough thing has always been, what should the shape of this browser be?

它到底是只给智能代理用的，还是别的？

Like is this something that is only for the agent, right?

就像你打开 Chrome，用它做你该做的事。

That like you've got Chrome, you open Chrome, you know, do your thing in Chrome.

如果你让桌面应用打开浏览器，它能立刻启动并直接控制，没有 Playwright 那样的延迟。

If you ask the desktop app, it opens up this browser that it can control really quickly, doesn't have the latency of playwright, whatever.

但另一方面，我们是不是想说，这个应用啥都能干，希望你把它当日常浏览器用？

But that, you know, or are we trying to say, this app is for everything and like, we want you to use this as a browser.

这两种思路各有优劣。

And those have a lot of trade offs.

这条路其实没多少人走过，对吧？

It's not super, you know, well traveled path, right?

大多数浏览器都是作为顶层应用存在的。

Like most browsers are browsers at the top level.

它们都有标签页。

They've got browser tabs.

这就带来很多枯燥但棘手的问题，比如快捷键怎么处理。

This creates a lot of really boring but tedious problems like keyboard shortcuts, right?

我们是要把快捷键映射到 Versus Code，还是 Chrome，还是我们自己的系统，又或者是 Linear？我们希望用户能形成一套通用的肌肉记忆，但市面上这些产品各有各的结构和操作方式。

Are we trying to do key mapping to Versus code or to Chrome or to our own thing or to linear or like, we wanna have some sort of muscle memory that carries over but got all these things that have shape, like, sub shapes of different products out on the market.

那我们到底该怎么做？

What do we do?

这恰恰说明了这个应用的难度有多高——你得让它同时适合从完全没编程经验的普通用户，到像彼得这样的高级用户。

And this just highlights how extra challenging this app is, where you have to allow it to work for somebody that's never built anything from like the more basic user to like power, like Peter.

对。

Yeah.

OpenAI 也想用它来写代码。

OpenAI trying to code with it.

我不觉得彼得会愿意用这个应用。

I'm not convinced I'm gonna get Peter to use the app.

我觉得他可能是最后一个还坚持用终端的人了。

I think he might be the last Is it terminal?

真不是顽固不化。

Not really hold out.

好吧。

Okay.

但我还是会继续尝试。

But I'm gonna I'm gonna keep trying.

让我退一步，从整体来看，你刚才说的这些。

Okay, let me zoom out for a moment and talk about kinda the big picture, and for you're taking all this.

Codex 的最终目标是什么？

What's the what's kinda the vision for Codex?

它将来会走向哪里？

Where does this go?

它最终会变成什么样子？

What's it gonna look like?

一年两年？还是十年后？

I don't know, a year or two, ten years?

我们最初做的是 Codex 的命令行工具，后来才决定开发这个应用。

We had Codex as a CLI, And then we decided to build this app.

说实话，我们一开始对这个应用还挺犹豫的，但对它能成为一款开发者工具充满信心。

And you know, we were we were a little uncertain about the app and but had a lot of conviction in what it could be as to start a developer tool, right?

它不会是个完整IDE，而是想打造一个刚刚好的界面——有点像聊天机器人，但又不止于此，你能看到代码，但不会让你直接修改。

And it wasn't gonna be an IDE, was gonna be this right sized surface where it was like sort of a chat bot but it was more than that and you could see the code but we weren't gonna let you edit the code.

在OpenAI，今年一月和二月发生了一件特别有意思的事。

There's a really interesting thing that happened at OpenAI in January and February.

那会儿我们还没正式发布Codex应用，但已经开始内部使用了。

And it was before we actually released the Codex app, which is that we'd started to dog food the Codex app.

我们发现，工程师和研究人员的使用体验逐渐清晰，内部已经找到了不错的契合点。

And what we were finding, like we were converging on some pretty clear internal PMF on engineering, right, and research workflows.

他们特别兴奋。

They were thrilled.

都爱上了这个工具。

They were loving it.

我们就想，行，先把质量提上去，再对外发布。

We were like, alright, we just gotta get the quality bar up before we release it to the world.

我们坚信，这一定会成事。

We're convinced that this will be a thing.

但后来在公司里，我们又尝试了几个其他的工作流程，心想：这个Codex项目在编程助手方面确实有门道。

But then at the company we spun up a few other workflows to say, hey, like this Codex effort is onto something with these coding agents.

市场、公关、财务、法务，几乎每个部门的人都在用这个Codex应用，尽管它对这些人来说根本就是个反直觉的产品。

And we have people from marketing, from comms, from finance, from legal, from basically every discipline who are using this Codex app even though it is actively hostile to these people, right?

这就像是硬要给他们看代码。

It is like trying to show them code.

还得让他们批准在系统里运行RG，诸如此类——全是些完全不适合他们的功能。

It's trying to ask for approval to run RG on the, you know, it's like it's doing all of these things that are actively not the right product surface for them.

那我们为什么不把其他界面也加上Codex呢？

So why don't we take our other surfaces and add Codex to them?

比如，把Codex加到ChatGPT的桌面应用里。

Like let's add it to the ChatGPT desktop app.

再比如，加到Atlas浏览器里。

Let's add it to the Atlas browser, right?