We Built Grok Bot. Here Are Our 14 Best Bots | Peng Zheng & Lauren Tan

频道: Peter Yang
视频: https://www.youtube.com/watch?v=xZ5TEaleUdg
原文语言: en
统计: 共 137 轮 · Peng 26 · Peter 34 · Lauren 54

[0:00] Peng

I will have my PM design like [music] three individual bots edited them into a single group chat and then they'll start to discuss and ideulate about the sync. I [music] will do the first 5% and I'll create like one key frame and then ask it to scale [music] it into the whole flow and make it like an end to end range.

我会把我的 PM、设计……总共三个独立的 bot 拉进同一个群聊，然后它们就会开始围绕这件事讨论、构思。前 5% 我自己来做，比如先做出一张关键帧，然后让它把这张扩展成整个流程，做成端到端的一整套。

[0:17]

It lets me orchestrate like really massive agent swarms. Sometimes I actually don't even look at the PR until after it's landed and then I'm like, "Oh, okay. Yeah, that looks good." I [music] think it ultimately comes back to trust, right? How much trust are you willing to give your agent or your bot to be able [music] to do a task for you? And I think one thing I always try to encourage people to do is to actually spend the time to sort of observe the bot or agent doing the work. All right. Hey everyone. My guests today are Pen and Lauren who are members of the design and technical staff on the Grockbot team. really excited to see their Grockbot setups for work and personal tasks and also how to use Grockbot to design and build the actual product and also the most surprising bots that they see the community making. Welcome both.

Lauren： 它让我能编排规模非常大的 agent 集群。有时候我甚至等 PR 合入之后才去看，然后心想：「哦，好。嗯，看着没问题。」我觉得归根结底还是信任的问题，对吧？你愿意给你的 agent 或 bot 多少信任，让它能替你把一件事办了？我一直鼓励大家做的一件事，就是真的花点时间，去观察 bot 或 agent 是怎么干活的。

Peter： 好。大家好。今天的嘉宾是 Peng 和 Lauren，他们分别是 Grok Bot 团队的设计成员和技术成员。非常期待看看他们在工作和个人事务上的 Grok Bot 配置，看看怎么用 Grok Bot 来设计和构建这个产品本身，还有他们在社区里见过的最让人意外的 bot。欢迎两位。

[1:03]

Thanks for having us.

谢谢邀请。

[1:04]

Thank you.

谢谢。

[1:04]

Very excited.

很期待。

[1:05] Peter

Yeah, very excited. I I think Grockbot is very exciting. I think uh it's the first persistent, you know, personal computer product that I've seen and I use it every day myself. But why don't we start with this? Um maybe you guys can show us kind of your personal Grogbot setup and a few of the bots that you're really proud of. Maybe let's start with you, Pen.

对，很期待。我觉得 Grok Bot 真的很让人兴奋。我觉得它是我见过的第一个常驻型的——你懂的——个人电脑产品，我自己每天都在用。不过我们先从这个开始吧：你们可以给我们看看各自的 Grok Bot 配置，还有几个你们特别得意的 bot。要不先从你开始吧，Peng。

[1:23] Peng

Yeah. So um I'd like to share a couple of bots I have. So basically I'm using Grabbot for almost everything um in the digital world I'm having. And at a at a high level I categorize them by work and life. Um and then I have like those most frequently used bots pen here. Um, chief of staff is my go-to bot. If I don't know which one should take the task, then I'll send it to chief of staff. Uh, and one thing is you can almost ask it to do anything in the digital world. Um, I was asking it to buy 3D printing filament um, couple days ago and it does not only buy it for me. I also ask it to add it to the inventory list. So I have this notion database to track all my inventory different colors of the filament and once they once they bought it it also like update the list for me.

好。那我来分享几个我的 bot。基本上，我数字生活里的几乎所有事情都在用 Grok Bot。大的层面上，我按工作和生活给它们分了类。然后最常用的那几个 bot 我置顶在这儿。chief of staff（幕僚长）是我的首选 bot，如果我不知道某个任务该交给哪个 bot，就直接发给 chief of staff。还有一点，数字世界里的事，你几乎什么都能让它做。前几天我让它买 3D 打印耗材，它不光帮我买了——我还让它把东西加进库存清单。我有个 Notion 数据库，记着我所有的库存、各种颜色的耗材，它买完之后还会顺手帮我把清单更新了。

[2:19] Peng

Um I just discovered like the benefit of trying different tasks together is really powerful because another example is like let's say I want to sell my DJI mic on Facebook marketplace. Um I just send it a photo and then tell it do a couple things. So like sell something, list something on Facebook. Marketplace sounds easy, but actually there are a lot of steps like you need to check the latest price for the model on the official website. I wanted to find out what's the lowest price offered in my area, like who I'm competing with. I wanted to learn like how I write listing descriptions uh in the past and also post this new listing. And then if nobody's interested in it, automatically lower the price by $5 every week. So you see like it's like a series of tasks that I dedicate in one single prompt and it will go execute the task and it's very interesting that it know like when it should write the description it consult my writer bot so it message the writer bot it ask like how I wrote uh marketplace listing description in the past and then it will able to do the research it'll be find out like how other people are selling it and then eventually pull up my list. So in the past we may do like all those five steps manually by ourselves but now not only can delegate individual task to it but also like ask it to train the tasks for you. Um my bot is replacing my email. So like I would do things like hey what are the action items in my inbox today? Uh my email is still out of manage and like for things I I identified that can do it for me. For example, I need to pay my friend money for dinner. Uh I'll just ask it to do so. And if it doesn't have access, um it will prompt this signin page for me and I can take over and sign in on its computer and it will do it for me. Uh it's replacing my calendar. So if I'm having a coffee chat with somebody now, like basically any edits, any event creation, find booking like book

我刚发现，把不同任务串起来一起做，威力真的很大。再举个例子，比如我想在 Facebook Marketplace 上卖掉我的 DJI Mic，我只要发一张照片给它，然后让它做几件事。在 Facebook Marketplace 上卖个东西、挂个商品，听起来简单，其实有很多步骤：得去官网查这个型号的最新价格；我想找出我这一带最低的报价是多少，也就是看看我在跟谁竞争；我想学一下我以前是怎么写商品描述的，还要把这条新商品发出去；然后如果没人感兴趣，就每周自动降价 5 美元。所以你看，这是一连串任务，我用一条 prompt 全部委派出去，它就会去执行。很有意思的是，它知道到了该写描述的时候，要去请教我的写作 bot——它会给写作 bot 发消息，问我以前是怎么写 Marketplace 商品描述的；然后它还能做调研，搞清楚别人是怎么卖这个东西的，最后把我的商品挂上去。以前这五个步骤我们可能都得自己手动做，但现在不光能把单个任务委派给它，还能让它帮你把这些任务串起来。我的 bot 正在取代我的邮箱。比如我会问：「嘿，我今天收件箱里有哪些待办事项？」我的邮件还是多到管不过来，对于那些我认定它能替我办的事——比如我得把吃饭的钱转给朋友——我就直接让它去办。如果它没有权限，它会给我弹出登录页，我可以接管，在它的电脑上登录，然后它就替我办完。它也在取代我的日历。比如我现在要跟某人约个咖啡聊天，基本上任何修改、任何新建日程、找时间预订，比如订

[4:23]

conference rooms, anything I just put a screenshot and then let it do it. let it like add the events to the calendar, modify it, um stuff like that. Um I noticed you have a Okay, so you have a Okay, let's let's go up to your ping bots. So you have a designer and a writerbot pinned. So designer So the writer is like your writing style and designer is like Yeah, maybe can quickly walk through those.

Peng： 会议室，什么都行，我只要丢一张截图过去，让它去办就行——让它把活动加进日历、修改之类的。

Peter： 我注意到你有个……好，你有个……好，我们往上翻到你置顶的那几个 bot。你置顶了一个设计师 bot 和一个写作 bot。所以设计师……写作 bot 大概就是照着你的写作风格来写，设计师 bot 就是……嗯，要不你快速带我们过一下这几个？

[4:48] Peng

Yeah. So, this is a demo account that I don't have like everything here, but it's the exact same setup I have in my like in my own account. Um, for like those three are the most frequently used like chief of staff going to manage all other bots. And the designers like I stayed in design tools every day and then designer is my go-to bot. I'll have it like brainstorm with me or like ask it to um spin off different design options for me. Um, and I'll just like discuss with it and not only like work with me on it, but also execute the ideas. And then for the writerbots, um, currently like I I'm not an English native speaker. And then a lot of emails and Slack messages, I would have like the writerbot to polish it before I send it. And we have the Slack and email chat widget. So you literally see um the draft that inline here that you can you can review and send and decide how to modify etc.

好。这是个演示账号，不是所有东西都在这儿，但跟我自己账号里的配置一模一样。这三个是我最常用的：chief of staff 负责管理其他所有 bot；然后是设计师 bot——我每天都泡在设计工具里，设计师 bot 就是我的首选。我会让它跟我一起头脑风暴，或者让它帮我衍生出不同的设计方案。我会跟它讨论，它不光跟我一起琢磨，还会把想法落地执行。然后是写作 bot：我英语不是母语，很多邮件和 Slack 消息，我都会先让写作 bot 润色一遍再发。我们还有 Slack 和邮件的聊天小组件，所以你能直接在这里看到内嵌的草稿，可以审阅、发送，决定怎么改，等等。

[5:45] Peter

Got it. This episode is brought to you by Granola. If you're in backtoback meetings, you know how much work it is to take notes live and clean them up afterwards. That's why I love Granola, the best AI [music] meeting notes app in the market. Here's how I use it. Granola automatically takes notes during a meeting and I can add my own notes, too. After the meeting ends, I use a Granola recipe to extract clear takeaways and next steps in the exact format that I want. Then I can just share her notes directly in Slack with my [music] colleagues or even get Granola to share her notes automatically. Honestly, of all the AI apps that I use, Granola is the one that saves me the most time. Try it now at granola.ai/pater and use the code Peter to sign up and get three months free. That's granola.ai/peter. Now, back to our episode. You also have a channel like me now app or something where you have a bunch of bots talking to each other.

明白了。本期节目由 Granola 赞助。如果你整天一个会接一个会地开，就知道现场记笔记、会后再整理有多费劲。所以我很喜欢 Granola，市面上最好的 AI 会议笔记应用。我是这么用的：开会时 Granola 会自动记笔记，我也可以加上自己的笔记。会议结束后，我用 Granola 的一个 recipe（预设的整理模板），按我想要的格式提炼出清晰的要点和下一步行动。然后可以直接把笔记分享到 Slack 给同事，甚至让 Granola 自动分享笔记。说实话，在我用的所有 AI 应用里，Granola 是最帮我省时间的那个。现在就去 granola.ai/peter 试试，注册时用优惠码 Peter，可以免费用三个月。网址是 granola.ai/peter。好，回到正题。你还有一个频道，好像叫 meow app 什么的，里面有一堆 bot 在互相聊天。

[6:37]

Yeah. So [laughter] what is that?

对。所以（笑）那是什么？

[6:41] Peng

Once you have bigger tasks that um you want a different bots with different rows have their own perspective and contribute to it.

就是当你有更大的任务，想让担任不同角色的不同 bot 各自带着自己的视角来出力的时候。

[6:50] Peng

Um this is an example like if let's say we are building an app that can translate cat meow to English and I want to start review this idea from different perspectives. So I will have my um PM design like three individual bots addit them into a single group chat and I will throw this prompt and then they'll start to discuss and ideulate about this thing and it's interesting to see like how they think about the idea how they debate with each other. Um if you want a specific bot to reply to you just simply at mention it and only that bot will reply to you. uh if you just stand a message in the group chat just like how you message your friends in group chat they will review it and then they will reply they will decide if they should reply uh from their own end go

举个例子，假设我们要做一个能把猫叫翻译成英语的应用，我想从不同角度来审视这个想法。那我就把我的 PM、设计……总共三个独立的 bot 拉进同一个群聊，把这个 prompt 丢进去，然后它们就开始围绕这件事讨论、构思。看它们怎么思考这个想法、怎么互相辩论，挺有意思的。如果你想让某个特定的 bot 回复你，直接 @ 它就行，只有那个 bot 会回你。如果你只是在群聊里发一条消息，就像你在群里给朋友发消息那样，它们会看一下，然后各自决定要不要回复。

[7:38] Peter

yeah this is really interesting I don't think this is easy to do in some of the other AI apps right because like keeping the context clean and having all these bots that have their own context chat with each other is like pretty unique

对，这真的很有意思。我觉得这在其他一些 AI 应用里不太容易做到，对吧？因为既要保持上下文干净，又要让这些各有独立上下文的 bot 互相聊天，这挺独特的。

[7:48] Peng

yeah it's um so there's like two angles you can think about the problem like some people manage their tasks and jobs by individual sessions, task and projects.

对。这个问题可以从两个角度来看：有些人是按一个个独立的会话、任务和项目来管理自己的任务和工作的。

[8:01] Peng

Um, and we are creating bots as more persistent named um, agents that have their own identity, memory system, had their own tools. U, it's really like abstract what's reusable into the row and then you can have them work together by creating separate chats and group chats for them. like even I have multiple projects with the same designer bot I can create another chat with it like for a different project right and then I can do this repeatedly for different projects I'm managing because one thing we noticed is a lot of time uh in AI agent product those chat are very ephemeral they're almost disposable like you don't care like maybe you only care like top three to five that you actively working on and then for the chats like you happened two weeks ago like you never go back. So I think this is a great way to like balance the two that for whatever reusable the capability the memory then you kind of abstract them into different rows and then for individual projects that has the end has the deadline and then you can create this dedicated sessions and work with your bots in it

而我们是把 bot 做成更常驻、有名字的 agent，它们有自己的身份、记忆系统，也有自己的工具。说白了就是把可复用的东西抽象成一个角色，然后你可以给它们建单独的聊天和群聊，让它们一起干活。比如同一个设计师 bot 我好几个项目都在用，我可以再跟它开一个聊天，专门用于另一个项目，对吧？我管的不同项目都可以这样反复来。因为我们注意到一点：很多时候，在 AI agent 产品里，那些聊天都非常短命，几乎是用完即弃的，你根本不在乎——你可能只关心手头正在做的三到五个，至于两周前的那些聊天，你再也不会回去看。所以我觉得这是在两者之间取得平衡的好办法：凡是可复用的能力、记忆，就抽象成不同的角色；而那些有终点、有截止日期的单个项目，你就可以建专门的会话，在里面跟你的 bot 一起干活。

[9:13]

and but but just like a human the designer bot like if I talk to designer bot for mail app and bark app it knows about both right in its context. Yes.

Peter： 那……但就像人一样，这个设计师 bot——比如我跟设计师 bot 聊 meow app，又聊 bark app，它的上下文里两个都知道，对吧？

Peng： 对。

[9:21] Peter

Or is it separate?

还是说是分开的？

[9:22] Peng

Yes. So when a bot is working in a group chat, you will able to see like the work status has changed in its own DM. So it's the same entity. It's the same bot working on your task.

对。当一个 bot 在群聊里干活时，你在它自己的私聊里也能看到它的工作状态变了。所以它是同一个实体，是同一个 bot 在处理你的任务。

[9:33] Peter

Got it.

明白了。

[9:33] Peng

Yeah. And then I have other use cases I think is pretty fun like grabbot is replacing my podcast tool right now because I listen to a lot of podcast but sometimes when I read an article I want to listen to it when I'm driving or taking shower. So I'll just simply send this link to my podcast bot and it will translate it into a podcast uh that I can listen anytime I just post.

嗯。然后我还有些别的用法，我觉得挺好玩的。比如 Grok Bot 现在正在取代我的播客工具，因为我听很多播客，但有时候读到一篇文章，我想在开车或洗澡的时候听。那我就直接把链接发给我的播客 bot，它会把文章转成一集播客，我随时都能听，发一下就行。

[9:58] Peng

So it's very it's very um interesting that you see it become like automate auto ultimate translator for different file formats.

所以挺有意思的，你会看到它变成了一个自动……终极翻译器，在各种文件格式之间来回转换。

[10:07] Peng

Um and also I like to share this personal project like the pipeline I set up. Um, the story was I took a photo of a produced shop in San Francisco, Chinatown like two weeks ago and when I was waiting for the bus and then I turned it into like a 3D clay miniature uh using AI and I thought, "Oh, this was fun." and I put it onto my personal website and it was very minimal website and I put it here and then I wrote the line that lasting at a JBC market in Chinatown and then I thought oh this could become a checkin app because for a personal website like you put it up there and people visit it once and then nobody came back again like I don't even visit it that often so I had this idea of making it an internet check-in app uh so every time I check in it can just automatically post the location and post the shop or store whatever um up there. So I started to talk to Grogbot about this idea and eventually I was able to land it on this pipeline that if I texted it uh a image of the place I've been and also like who am I with the social handle of the person it will be able to generate um two images in day and light mode and also remove the background um and then eventually post it to my website um something like is so

另外我还想分享一个个人项目，就是我搭的一条流水线。事情是这样的：大概两周前，我在旧金山唐人街等公交的时候，拍了一家蔬果店的照片，然后用 AI 把它做成了一个 3D 黏土微缩景观。我心想：「哦，这还挺好玩的。」就把它放到了我的个人网站上——那是个非常极简的网站，就放在这儿，然后我写了一行字：「最近出没于唐人街的 JBC market」。然后我想，哦，这可以做成一个打卡应用啊。因为个人网站嘛，你把它挂上去，人家来看一次，之后就再也没人回来了，连我自己都不常去看。所以我就想把它做成一个网上打卡应用，这样我每次打卡，它都能自动把地点、那家店之类的发到网站上。于是我开始跟 Grok Bot 聊这个想法，最后落地成了这样一条流水线：只要我给它发一张我去过的地方的照片，再告诉它我跟谁在一起、附上那个人的社交账号，它就能生成日间和夜间两种模式的两张图，还会把背景去掉，最后发到我的网站上。大概就是这样，所以……

[11:39] Peter

Oh, nice. Okay.

哦，不错。好的。

[11:41] Peng

It was it was uh it was very [laughter] whimsical, fun, like personal project. Um but I think it's a different uh perspective to think about it is now I'm using Grubbot as like the main interface to to to host things instead of like having build a portal or CMS system and manually upload it in the in the back end. And a lot of it is baked in

这其实就是个很随性（笑）、很好玩的个人项目。不过我觉得可以换个角度来看：现在我是把 Grok Bot 当成发布内容的主入口，而不用去搭一个门户网站或 CMS 系统、再手动到后台上传。而且很多步骤都已经内置在

[12:08] Peng

um

嗯……

[12:09] Peng

baked in baked in this pipeline that I have this bot just go fetch the latest playbook on this private repo uh and then follow the step and then like generate the image clean it up and then post it.

内置在、内置在这条流水线里：我让这个 bot 去一个私有仓库拉取最新的 playbook，然后按步骤走，生成图片、把它清理干净，再发布上去。

[12:21] Peter

Okay. So, so people can see all the places you've been to and and the typical designer you think about both day and light mode. I would not think about that. Yeah, that was good. Yeah. And

好，这样大家就能看到你去过的所有地方了。而且真是典型的设计师，日间模式、夜间模式都考虑到了，换我根本想不到这一层。对，这个很棒。嗯，然后

[12:31] Peng

yeah, so that's example that you can use grabbot in a creative way almost think it as like a building module in the thing you want to build and also the best part is it can guide you step by step towards that goal.

对，这就是一个例子：你可以很有创意地用 Grok Bot，几乎可以把它当成你想搭建的东西里的一块积木。而且最棒的是，它能一步一步引导你走向那个目标。

[12:43] Peter

And before we move on to Lauren um your designer bot so are are you still like drawing boxes in Figma or you tell your bot to do everything and you just re review?

在转到 Lauren 之前，再问一下你的设计师 bot：你现在还在 Figma 里画框吗？还是全都交给 bot 去做，你只负责审？

[12:51] Peng

We so as xaii we do we we design in whatever tools we need. Uh we use Figma, we use canvas based tools, we have our own prototype environments and sometimes we also submit our own PR.

我们……在 xAI，我们做设计时需要什么工具就用什么工具。我们用 Figma，也用基于画布的工具，还有我们自己的原型环境，有时候我们也会自己提 PR。

[13:04] Peng

Um so the designer bot is really helpful to bounce in ideas with me. I also make quick changes um because it connect the MCP Figma MCP through the connector. So it it will be able to like directly

所以设计师 bot 特别适合陪我碰撞想法。我也会用它快速改稿，因为它通过 connector（连接器）接上了 Figma MCP，所以它能直接

[13:18] Peng

change frames and update the design and use design token. Um I have this skill that the designer bot is able to use that so it know like how my Figma uh file setup is what's my design system and when to use the right like colors typography spacing um when it created design for me.

改 frame、更新设计、使用 design token。我有一个设计师 bot 能用的 skill，这样它就知道我的 Figma 文件是怎么组织的、我的设计系统是什么，以及它给我出设计的时候，什么时候该用对的颜色、字体排版和间距。

[13:39] Peter

Okay. So maybe it can do like the first 80% you can manually tweak stuff or something.

好，所以它大概能先做完前 80%，剩下的你再手动微调之类的。

[13:43]

Yeah.

对。

[13:44] Peng

Yes. Yeah. Or sometimes like my workflow has become I will do the first 5%.

是的，对。或者有时候，我的工作流已经变成了：我只做前 5%。

[13:49] Peng

Uh I would create like I would build the system and I'll create like one key frame and then ask it to scale it into the whole flow and make it like an end to end uh screens.

我会先把系统搭好，做出一个关键帧，然后让它把这个扩展到整个流程，做成端到端的一整套界面。

[14:00] Peter

Oh, awesome. You should like open source your design skills like Lauren did with her PP stack.

哦，太棒了。你应该把你的设计 skill 开源出来，就像 Lauren 开源她的 P-Stack 那样。

[14:05] Peng

Yeah. Yeah. Interesting. Like we have this

对对，有意思。我们有这种

[14:07] Peng

this skills follow our names like Lauren [laughter] mode and I have a pen mode.

以我们名字命名的 skill，比如 Lauren mode（笑），我也有一个 Peng mode。

[14:14]

Pen mode. Yeah. Think exactly.

Peng mode。对，我觉得正是这样。

[14:16] Peter

Awesome. Pen, thanks so much for sharing all this. This is super cool. Lauren, thanks for waiting. I would love to see how your setup is, too. It's probably pretty different from

太棒了。Peng，非常感谢你分享这么多，真的超酷。Lauren，久等了，谢谢。我也很想看看你的配置，估计跟……挺不一样的

[14:23] Lauren

very different. Yeah. Uh, I wanted to share I guess more of the engineering use cases. So, one thing I've been recently working on is adding better support for Omachi, uh, which is a Linux distribution that DHH and a couple of other really awesome open- source folks have been working on. Um, and I've been using Grockbot to essentially kind of control uh this is like a this is a special build of Omachi which runs in a virtual machine on your Mac.

非常不一样。对。我想分享的更多是工程方面的用例。我最近在做的一件事，是给 Omarchy 做更好的支持——Omarchy 是 DHH 和另外几位非常厉害的开源开发者一直在做的一个 Linux 发行版。我一直在用 Grok Bot 来……基本上就是去控制——这是 Omarchy 的一个特别构建版，可以跑在你 Mac 上的虚拟机里。

[14:55] Lauren

Uh, which is still like very alpha but it works pretty well.

它还处在非常早期的 alpha 阶段，不过跑得挺好。

[14:58] Lauren

Uh, but it's really powerful because I can actually get my agent to literally like script and control the VM. So uh what better way to build support for you know this particular flavor of Linux than to have graphbot actually be able to open it on my computer uh and you know like interactively install it like it installed the graphbot for me. You can see like it's actually doing that right now.

但这真的很强大，因为我真的可以让我的 agent 直接写脚本来控制这台虚拟机。所以，要给这个特定的 Linux 发行版做适配，还有什么办法比让 Grok Bot 直接在我电脑上把它打开、再交互式地完成安装更好呢？比如它就自己帮我把 Grok Bot 装上了。你看，它现在其实就正在做这件事。

[15:23]

Okay. So graphbot is on its own computer but it's controlling your computer through omachi. So yeah, so Grapa has its own computer. Uh but it also has a local execution demon I guess that runs on your computer. So if you if you tell it to like another [clears throat] thing I I often actually do is this I have this agent called crumb and I often just like run out of hard disk space. So I'll just tell it like hey go through my hard disk and like you know find where is all the space going and it just goes off and like find stuff and to delete it. Uh but yeah, it has like full access to your computer and uh I mean depending on the permissions you give it, but it can uh you know then in this case for me because I wanted to try out Omashi and you know fix some bugs or uh work on the experience a little bit for Linux, it was very powerful to be able to actually have the bot control the virtual machine itself, the Linux virtual machine on my computer. uh cuz yeah that's like one thing I always talk about is like you know the importance of actually uh having the agent be able to verify its own work. So like this is one example where the agent can verify it work on your computer in addition to of course like you know it has its own computer uh that you can you know take a look and and see what it's doing. Uh but uh yeah that's like one of the the main use cases I have uh is is doing like a lot of engineering work uh on on graph.

Peter： 好，所以 Grok Bot 在它自己的电脑上，但它通过 Omarchy 在控制你的电脑。

Lauren： 对，所以 Grok Bot 有它自己的电脑，但它在你的电脑上也跑着一个本地执行守护进程（daemon），大概可以这么叫吧。所以如果你让它……比如我其实经常做的另一件事是这样的：我有个叫 Crumb 的 agent，而我经常硬盘空间不够用，我就跟它说：「嘿，把我的硬盘过一遍，看看空间都被什么占掉了」，它就自己跑去把那些东西找出来，然后删掉。不过对，它对你的电脑有完全的访问权限——当然，这取决于你给它开了哪些权限——但它可以……在我这个例子里，因为我想试试 Omarchy，顺便修一些 bug，或者稍微打磨一下 Linux 上的体验，能让 bot 直接控制虚拟机本身，也就是我电脑上的那台 Linux 虚拟机，这一点非常强大。因为这也是我一直在讲的：让 agent 能够验证它自己的工作，这件事非常重要。这就是一个例子，agent 可以在你的电脑上验证它的工作；当然除此之外，它还有自己的电脑，你随时可以去看看它在做什么。不过对，这就是我的主要用法之一：在 Grok Bot 上做大量的工程工作。

[16:53] Lauren

Um another thing I manage is like my own personal stack of plugins Pstack. Um so all of the work I do on this plug-in basically happens in here. Um so uh you know I'll kick off cloud agents cursor cloud agents uh which um will um essentially do eval for me. So in order to like make a change to my set of skills that I can trust that the skill the improvements are actually going to you know result in better performance uh I have uh like an eval playbook in PSA that teaches the agent how to actually like it'll spawn a couple of sub aents of different models. It'll try out the prompt. Uh it will, you know, make sure that uh the you know the intent the goal that we actually set out to achieve was achieved and then only if it does achieve it, it will actually uh you know land that that pull request.

嗯，我还在管理另一样东西，就是我自己的一套个人插件合集，叫 P-Stack。我在这个插件上做的所有工作，基本都是在这里完成的。我会启动 cloud agent——就是 Cursor 的 cloud agent（云端 agent）——基本上就是让它们替我跑 eval（评测）。为了让我对这套 skill 的改动是可信的，也就是确信这些改进真的会带来更好的表现，我在 P-Stack 里放了一份 eval playbook，教 agent 具体怎么做：它会拉起几个用不同模型的子 agent，把 prompt 实际跑一遍，确认我们当初想达成的意图和目标确实达成了；只有真的达成了，它才会把那个 PR 合入。

[17:53] Lauren

Um yeah, so that's uh quite a lot of the engineering work that I do.

嗯，对，这就是我做的工程工作里相当大的一部分。

[17:59] Peter

Okay. And and I noticed your your your Grock bots have very different names from Pence Grock. They're more [laughter] they have more they have more personality than just, you know.

好。然后我注意到，你在 Grok Bot 里的那些 bot，名字跟 Peng 的很不一样。它们更……（笑）更有个性，不只是那种……你懂的。

[18:09] Lauren

Yeah. I like I I I asked my So I have this bot called Dr. Eggbot, which I've also kind of shared uh on my socials. Uh but Dr. Eggbot is essentially my designer bot. Uh I mean, not a designer bot, but like a bot designer bot who designs bots for me. uh with all of the rigor, engineering rigor that I expect from the bots that I use. Um and uh I also get it to come up with food theme names. I don't know. I guess it's like an Asian thing. We we we love we love food so we got to call it, you know, call my botsman cute like Dau Fuku Gza. [laughter] So that's pretty fun.

对。我……我让我的……是这样，我有个 bot 叫 Dr. Eggbot，我在社交媒体上也分享过它。Dr. Eggbot 本质上是我的设计师 bot——我是说，不是做设计的 bot，而是设计 bot 的 bot，专门替我设计 bot，而且带着我对自己所用的 bot 所要求的全部严谨，工程上的严谨。然后我还让它给 bot 起食物主题的名字。我也说不清，大概是亚洲人的特点吧，我们就是太爱吃了，所以得给我的 bot 起些可爱的名字，比如 Daifuku、Gyoza（大福、煎饺）。（笑）挺好玩的。

[18:50] Lauren

Um and then similar to Pang, I also have a chief of staff. I have a personal assistant. um personal assistant helps me with things like lunch, my flights, uh uh booking hotels. Uh sometimes I I ignore her and then she gets a little bit sassy. [laughter]

然后跟 Peng 一样，我也有一个 chief of staff（幕僚长）。我还有一个私人助理，私人助理会帮我处理午饭、航班、订酒店这类事。有时候我没理她，她就会有点闹小脾气。（笑）

[19:08] Lauren

Uh then my chief of staff uh yeah like helps me with just all sorts of random things. But uh if you you see in my sidebar, I have a bunch of like engineer bots here. I have some social bots as well that manage my you know my social media.

然后我的 chief of staff 就帮我处理各种杂七杂八的事。不过你看我的侧边栏，这里有一堆工程师 bot，还有一些社交 bot，负责打理我的社交媒体。

[19:26] Lauren

Um so a lot of the work that I do goes through these two or these three agents I guess like either I'm you know talking to the chief of staff and my chief of staff isn't working with like my in lead uh matcha here uh or you know if it's Suki my personal assistant like she might you know um book flights for me. I haven't set up like a personal assistant team yet because I haven't I don't have that many I'm not that busy I guess that I need like you know a team of bots helping me there. So so far she's been she's been doing a really great job.

所以我的很多工作都会经过这两个，或者说这三个 agent 吧：要么是我在跟 chief of staff 聊，而我的 chief of staff 在跟我这边的工程负责人 Matcha 协作；要么如果是 Tsuki，也就是我的私人助理，她可能会帮我订机票。我还没搭私人助理团队，因为我没有那么多……我大概也没那么忙，还用不着一整队 bot 来帮我。所以到目前为止，她一直干得非常好。

[20:01] Lauren

Uh and then Dr. just comes up with bots for me whenever I'm like I'm starting to realize like oh you know I'm starting to use like one agent for a lot of things and that's kind of um uh I'll say like this means I'm adding like a lot of context to that agent that's

然后每当我开始意识到——哦，我开始拿一个 agent 干很多事了——Dr. Eggbot 就会给我造新的 bot。这有点……我会说，这意味着我往那个 agent 里塞了很多上下文，而这些上下文

[20:20] Lauren

you know kind of very diverse and very different

你懂的，五花八门、差别很大，

[20:24] Lauren

which can be like maybe confusing. So uh one of the things that Dr. Dr. Eggbot does as well. It's like has a routine uh where uh yeah like here where it will so this is like a routine that Dr. Eggbot will set up for you where it'll look through all of the bots that you have um and the transcripts that you've uh you know you've been chatting with them about uh and it'll propose things like

这可能会让它有点混乱。所以 Dr. Eggbot 还会做的一件事是，它有一个 routine（按计划定时自动运行的例行任务）……对，就像这里，这就是 Dr. Eggbot 会帮你设好的一个 routine：它会把你所有的 bot，还有你跟它们聊过的聊天记录都过一遍，然后提出一些建议，比如

[20:50] Lauren

uh you know this bot keeps

呃，就是，这个 bot 老是

[20:53] Lauren

keeps doing this thing that's not so good and then it'll suggest like some improvements whether it's skills or a new bot or uh adjustments to routines. Uh it also goes through your existing routines as well and make sure that um they're efficient because routines are actually quite uh if you have them run too often it can be quite expensive on your usage because it just keeps you know waking up the agent like if you do like I don't know every 10 minutes that's like a lot of wakeups. Uh

老是在做某件不太好的事，然后它会提出一些改进建议，可能是 skill、新 bot，或者对 routine 的调整。它还会把你现有的 routine 过一遍，确保它们够高效，因为 routine 其实挺……如果你让它们跑得太频繁，会很耗你的用量，因为它会不停地唤醒 agent。比如你设成——我也不知道——每 10 分钟一次，那唤醒次数就非常多了。呃

[21:23] Lauren

so um Dr. Abot does help a lot with that.

所以，嗯，Dr. Eggbot 在这方面确实帮了大忙。

[21:27] Peter

Okay. Well okay I'm going to give you a couple questions. So

好。嗯，好，我来问你几个问题。那么

[21:30] Lauren

sure

好啊。

[21:30]

for for Sukiebot. So you actually rely on to book flights and stuff like do you actually rely on to make like big purchases or what like what what's the most autonomy you've given it to? So uh yeah, I've actually booked my so I'm going to give a talk soon uh in London and Amsterdam as part of the compile on the road conference and so uh hopefully I don't leak my flight details but uh yeah like basically I booked the entire trip through Tuki um and this was actually quite I was actually quite impressed by how she did it because it was a very it was kind of a complex like multi-step flight. I don't know what you call it. Like not it's not a simple just like return flight from Orange County to London. It was like Orange County and then to like Denver and then to London and then London to Amsterdam and then Amsterdam back to back to my home.

Peter： 关于 Tsuki bot。你真的会靠它订机票这些事吗？比如，你真的会靠它去买大件吗？或者说，你给过它的最大自主权到什么程度？

Lauren： 嗯，对，我其实已经订了我的……是这样，我很快要去伦敦和阿姆斯特丹做演讲，是 compile on the road 大会的一部分。希望我别把航班信息泄露出去——不过对，基本上整趟行程我都是通过 Tsuki 订的。她的做法其实让我挺惊艳的，因为这是一趟挺复杂的多段航程，我也不知道这该叫什么。不是那种简单的橙县（Orange County）到伦敦往返，而是先从橙县飞到丹佛，再飞伦敦，然后伦敦到阿姆斯特丹，最后从阿姆斯特丹飞回家。

[22:24] Peter

Yeah.

嗯。

[22:24] Lauren

Uh and I booked it entirely through uh like we we use this uh what's it what's called Navan, I guess.

而且我是完全通过……我们用的那个，叫什么来着……叫 Navan 吧，好像是。

[22:33] Peter

Okay. Corporate thing. Yeah.

好，公司用的那种东西。嗯。

[22:34] Lauren

Yeah. It's like a corporate thing to like book flights for you. Uh but it did all of the work in in finding the best you know legs and the best uh the best class of flight that I could get away with uh the best timing you know and it also figured out all the hotels for me uh which were in close proximity to the event.

对，就是公司用来帮你订机票的那种平台。但所有活儿都是它干的：找最好的航段、在规定允许的范围内能订到的最好舱位、最合适的时间；它还帮我把所有酒店都找好了，都离会场很近。

[22:52] Lauren

So that was like a really it did it all entirely autonomously. Uh I just I I just gave it a link to a Slack message and with some of the details about the conference and said book some hotels and flights for me and you know check in with me before you hit approve uh sorry if you hit book and then it came up with some proposals.

所以那真的是……整件事完全是它自主完成的。我只是给了它一条 Slack 消息的链接，里面有大会的一些细节，然后说：帮我订一下酒店和机票，在你点「批准」之前——抱歉，是在你点「预订」之前——先跟我确认一下。然后它就拿出了几套方案。

[23:12] Lauren

Uh and I just yeah I just hit yes and it just booked everything for me. So that was very very nice. It's probably the biggest yeah biggest purchase I've done through um through GOP

然后我只点了「是」，它就把所有东西都帮我订好了。这真的非常非常棒。这大概是我通过……通过 Grok Bot 做过的最大一笔消费了

[23:25]

I guess. Uh, okay. But but you still kind of got to like outpoint chat what he plans to book and stuff, right? Before you

Lauren： 我猜吧。

Peter： 呃，好。不过你还是得先过目确认它打算订什么之类的，对吧？在你……之前

[23:31]

Yeah, it wasn't uh Yeah, it wasn't to the point yet where, you know, I I guess I didn't want to have like some crazy redeye flight uh or or, you know, like be put in like the back of the plane. Uh so I Yeah, I I I wanted to approve it first. Yeah, I have a very similar bot uh that mons I'm I'm trying to book a vacation to Japan again and like you know you can set up like price alerts on Google flights but like the bot's more smart because it can find like other routes and stuff that I haven't set up. So it actually saved me like a bunch of money recently.

Lauren： 对，还没到那个程度……我大概就是不想被订上什么离谱的红眼航班，或者被塞到飞机最后几排。所以我……对，我想先自己批准一下。

Peter： 对，我也有个很像的 bot，专门盯着……我最近又在订去日本度假的行程。你知道，Google Flights 上也能设价格提醒，但这个 bot 更聪明，因为它能找到我没设置过的其他航线之类的。所以最近它还真帮我省了一大笔钱。

[24:01]

Yeah.

对。

[24:01] Lauren

Yeah, it definitely can. Yeah, [snorts] I also use it to book uh buy my uh personal things like I we're running out toilet paper for example. I'll just tell, hey,

对，它确实能。嗯，我还用它帮我买个人用品，比如家里卫生纸快用完了，我就直接跟它说：嘿，

[24:12] Lauren

get us reorder our Amazon order for for toilet paper and it just goes off and does that. There was a point actually like very early in the grabbot development where we were still kind of uh iterating on it that uh one of the topics that came up was your bot being able to text you. So, one of the things I actually added to Tuki a while back was this uh I basically gave her her own phone number through uh a third party service. Um, and then she would send so then she could actually text me because one thing I'm really bad about is remembering to eat lunch.

帮我们把 Amazon 上买卫生纸那单再下一次。然后它就自己去办了。其实在 Grok Bot 开发的很早期，我们还在反复迭代的时候，有个话题被提了出来，就是让你的 bot 能给你发短信。所以我前阵子给 Tsuki 加的一个东西就是——基本上是通过一个第三方服务，给了她一个自己的手机号。然后她就会发……这样她就真的能给我发短信了，因为我有件事特别不在行，就是记得吃午饭。

[24:46] Lauren

So I need the reminder uh and and then I often will sometimes just get I'll just sometimes have I don't know like do not disturb mode on on my on my Mac. Uh and I don't see the you know the the ping uh but then I get the text message and it makes a sound and like oh yeah I have to order lunch. Uh, and so I also taught uh, Tuki how to use Door Dash, of course, and you know, get me get me lunch. Uh, so I don't have to to worry about it so much.

所以我需要有人提醒。然后我经常会……有时候我 Mac 上开着勿扰模式，就看不到那条提醒，但短信一来会响一声，我就想起来：哦对，我得点午饭了。所以我当然也教会了 Tsuki 用 DoorDash，帮我把午饭点上，这样我就不用太操心这件事了。

[25:13] Peter

That's funny. Lunch is the highlight of my day, so I always look forward to [laughter] never forget my lunch.

真逗。午饭可是我一天里的高光时刻，所以我总是盼着……（笑）我从来不会忘了吃午饭。

[25:17] Lauren

Yeah. Yeah. Some I don't know. You just get too in the zone, you know, and then you just forget to eat.

对对。有时候……我也说不清。就是一进入状态就太投入了，你懂的，然后就忘了吃饭。

[25:22] Peter

Okay. So my second question is I noticed you have a lead and two engineer bots, right? So basically you

好。那我的第二个问题是——我注意到你有一个负责人和两个工程师 bot，对吧？所以基本上你

[25:28]

I have three. each each one works on a ticket or something or how did that dedicate responsibility to

Lauren： 我有三个。

Peter： 是每个各负责一张工单之类的吗？还是说你是怎么把职责分派给

[25:33] Lauren

um yeah so I actually mostly talk to my chief of staff and then my chief of staff talks to the engine lead uh but I set these bots these engineer bots up with Dr. Eggbot of course uh but one of the things I kind of drove uh I I told Dr. for eggat to do is to make sure that the end lead's job is really to you know break down tasks into smaller pieces and delegate and uh supervise other bots rather than do work on its own.

嗯，对，其实我主要是跟我的 chief of staff（幕僚长）对话，然后 chief of staff 再去跟工程负责人沟通。这些 bot——这些工程师 bot——当然都是我用 Dr. Eggbot 搭起来的。不过我着重推动的一件事，也就是我交代 Dr. Eggbot 去做的，是确保工程负责人的职责真的就是把任务拆成更小的块、分派下去、监督其他 bot，而不是自己动手干活。

[26:01] Lauren

So you can see in the the description that basically it has like even the the bot ID is in there uh but it tells it to you know never do work on your own and always uh delegate to those four bots. Uh and then it will it basically does that. Um and uh a lot of the work as well is done through cloud agents. Um so uh for example uh one of these bots will uh in order to do the work they will like create a cloud agent uh and then you can you know open that in cursor or on the web. Um,

所以你能在它的描述里看到，基本上——连 bot ID 都写在里面——描述里告诉它：永远别自己干活，一律分派给那四个 bot。然后它基本上也就是这么做的。另外，很多活也是通过 cloud agent（云端 agent）完成的。比如说，这些 bot 里的某一个为了把活干完，会去创建一个 cloud agent，然后你可以在 Cursor 里或者网页上打开它。

[26:40] Peter

okay.

好。

[26:41] Lauren

But yeah, this allows me to essentially make use of different models depending on the type of task. Uh, cuz you know, some tasks are are more straightforward and I just want like really fast uh speed. Uh, other tasks, you know, maybe require a bit more planning and a bit more reasoning and then I'll might adjust the reasoning level. So, cloud agents in graphbot give me a lot of like very granular control I guess. Um and also ensures that you know like my bot kind of stays I I kind of treat each of these engineers as more like like their like their supervisors in a way.

不过对，这样我基本上就能按任务类型用不同的模型。因为有些任务比较直接，我只想要速度非常快；另一些任务可能需要多一点规划、多一点推理，那我可能就会调整一下推理等级。所以我觉得 Grok Bot 里的 cloud agent 给了我非常细粒度的控制。而且也能确保我的 bot 能保持……我某种程度上是把这些工程师 bot 当成主管来看待的。

[27:15] Lauren

Uh we have like so we have like the lead that manages them but then they themselves try not to do work on their own. Uh so they always spawn um cloud agents for me. So this kind orchestrate yeah they they do all the message passing but lets me orchestrate like really massive agent swarms

我们有……就是说有个负责人管着它们，但它们自己也尽量不亲手干活，所以总是替我拉起 cloud agent。这样就……对，消息传递全是它们在做，但这让我能编排规模非常庞大的 agent 集群

[27:33] Lauren

uh to do like you know like if I need to tackle a very big project I can just talk to my chief of staff and say you know break this down you know I have a giant project uh maybe let's do some planning first in a notion doc and then it will start thinking about the phases of the of the work that it needs to do and then it'll delegate it all to the engine and who will then break it down even further and have the the bots kind of work on that.

来干活。比如说，如果我要啃一个非常大的项目，我只要跟我的 chief of staff 说：帮我拆解一下，我有个超大的项目，要不先在 Notion 文档里做点规划。然后它就会开始思考这项工作要分哪几个阶段，再把它全部分派给工程负责人，工程负责人会再进一步细拆，让那些 bot 分头去做。

[27:58] Peter

Okay. Then each one will come back with like a PR and stuff and the bots will re review it first. Is that kind of how it works?

好。然后每个都会交回一个 PR 之类的，再由 bot 先评审一遍。大致是这么运作的吗？

[28:02] Lauren

Uh yeah, essentially. Uh although for me, yeah. Uh one of the things that I've done a lot uh that I talk about a lot as well on X is the idea that like for example the Grothbot codebase, we've tried to make it very how do I say uh make it very agent friendly. The architecture and the codebase is very agent friendly. So, a lot of times it actually just lets me automerge my PRs. So, uh sounds kind of scary to say, but some sometimes I actually don't even look at the PR until after it's landed and then I'm like, "Oh, okay. Yeah, that looks good." Uh, [laughter]

呃，对，基本上是这样。不过对我来说……我做了很多、也经常在 X 上讲的一件事，就是比如 Grok Bot 的代码库，我们一直努力让它变得非常——怎么说呢——对 agent 非常友好。架构和代码库都对 agent 很友好。所以很多时候我其实可以直接让 PR 自动合并。说出来有点吓人，但有时候我甚至要等 PR 都合入了才去看，然后心想：哦，好吧，嗯，看着没问题。（笑）

[28:39] Peter

got it.

明白了。

[28:40] Lauren

Yeah.

对。

[28:41] Peter

Wow. Okay. So, so like you have like a software factory. Is that is that the buzz buzz word?

哇，好吧。所以你这等于是有了一座软件工厂？这是不是就是那个……那个流行词？

[28:46] Lauren

Uh, I like to call it the Michelin Kitchen because I think software factory. Yeah. Uh, yeah. I like to call it the Michelin kitchen because to me Michelin kitchen

呃，我更喜欢叫它「米其林厨房」，因为我觉得「软件工厂」……嗯。对，我喜欢叫它米其林厨房，因为对我来说，米其林厨房

[28:56] Lauren

you see like there's this uh when you when you think about a Michelin starred kitchen right you think about the quality right and the craft that goes into making those uh you know the the product right which is a a dish or a meal uh and there's still scale involved right like you know a kitchen a restaurant might have you know 50 seats or or or 100 seats uh so you need to be able to like produce quality at scale. So that's the aspirational vision I have is like I want to set a Michelin kitchen not just like you know a factory which uh interestingly I think a lot of people's like when you say factory they have this connotation like oh you know it's just like it's massured it's like yeah slob is like low quality

你看，一想到米其林星级厨房，对吧，你想到的是品质，对吧，还有做出那个产品——也就是一道菜、一顿饭——背后的手艺。而且这里面依然涉及规模，对吧？一个厨房、一家餐厅可能有 50 个座位或者 100 个座位，所以你得能在规模化的同时产出品质。这就是我心里向往的愿景：我想打造的是一间米其林厨房，而不只是一座工厂。有意思的是，我觉得很多人一听到「工厂」，就会联想到：哦，那就是批量生产的，就是……对，粗制滥造的垃圾（slop），就是低质量

[29:42]

but I I always try to aspire to do both right like how do we get quality at scale so yeah this is uh a lot of the work I've done on pack and like using rockbot together with it is in that line of you know how do we create that Michelin kitchen? Yeah, you should get that you should make like a concurge bar or something that talks like you know, hey, we curated this PR from the local farm and uh [laughter] talk like a talk like a vintage

Lauren： 但我一直努力两者兼得，对吧，就是怎么做到规模化的同时保证品质。所以对，我在 P-Stack 上做的很多工作，还有把 Grok Bot 跟它结合起来用，都是沿着这个思路：我们怎么打造出那间米其林厨房？

Peter： 对，你应该搞个……你应该做个礼宾 bot 之类的，说话就像：「您好，这份 PR 是我们从本地农场为您精选的」，然后……（笑）说起话来就像……就像那种复古的

[30:06] Lauren

Yeah. Yeah. Yeah. We need like artisal. Here's an artisal PR for you to to

对对对，我们需要那种匠心手作款。这是一份匠心手作的 PR，请您

[30:09]

review. So, so basically like Pun's job is just to like take your open source skills and somehow build into the product. Is that kind of what PP is trying to do or [laughter] No, I'm just kidding. It's like Yeah. Yeah. No, it's great. You have you definitely have a to system set up for

Lauren： 评审。

Peter： 所以说白了，Peng 的工作就是把你那些开源的 skill 想办法做进产品里？PP 想做的就是这个吗，还是……（笑）不不，开玩笑的。就是……对，对。不，这很棒。你确实已经搭起了一套系统，专门用来

[30:24] Lauren

Yeah. I think the interesting thing for me as well is like I found that with PAC and with like some of the so internally we uh we have like everyone has like not everyone but a lot of people have their own mode skills. So like potato mode uh is is is what is called uh in open source but then internally it's called lauren mode and you know like ping was sharing there's ping mode and like a lot of different you know engineers and you know designers PMs have their own mode skills um and I think this allowed us to really compound on some of the skills that we've already set up in the codebase. So yeah big a big motivation for uh uh of of the you know redesigning and architecting the graphbot codebase was so that everyone right could contribute at a high level not just engineers. So a lot of care has been taken into doing that. Um and of course yeah like people are kicking kicking these agents off in graphbot um as well as sometimes in cursor. Uh and the nice thing about cursor and graphbot plugins is that they interoperate. So you can install PAC for example on both uh grabbot as well as cursor and kind of reuse the the same skills.

对。我觉得对我来说还有一点很有意思：我发现有了 P-Stack，再加上一些……在内部，我们——几乎每个人都有——也不是每个人，但很多人都有自己的 mode skill。比如 poteto mode，这是它开源版的名字，在内部它叫 Lauren mode；还有像 Peng 刚才分享的，有 Peng mode；很多不同的工程师、设计师、PM 都有自己的 mode skill。我觉得这让我们能在代码库里已经搭好的那些 skill 之上真正做到复利式积累。所以对，我们重新设计、重新架构 Grok Bot 代码库的一大动机，就是让每个人都能在更高的层面上做贡献，而不只是工程师。所以我们在这上面花了很多心思。当然，大家会在 Grok Bot 里启动这些 agent，有时候也会在 Cursor 里启动。Cursor 插件和 Grok Bot 插件有个好处，就是两边能互通。比如你可以把 P-Stack 同时装在 Grok Bot 和 Cursor 上，复用同一套 skill。

[31:41] Peter

Interesting. Yeah, I didn't think about this. So So your bots can also call different skills, right? Like let's say I have like a right like me scale or some sort of scale. You just trigger it the slash.

有意思，这点我之前没想到。所以你的 bot 也能调用不同的 skill，对吧？比如说我有个「模仿我文风写作」的 skill，或者别的什么 skill，直接用斜杠就能触发？

[31:49] Lauren

You can Yeah, if you do like slash slash potato mode, you'll see that um it's uh you can use that skill there. Um, and you can always look at the marketplace as well. Um, I actually don't know if there are other plugins that are just skills, but PAC is one of them. Uh, but a lot of the plugins we have also come with skills. So, uh, in addition to like, you know, an MCP server. So, for example, like if you I don't know if you install the the X uh the X plugin for example, it includes I think one skill in addition to the tool. Um, and yeah, like so

可以。对，比如你输入斜杠，/poteto mode，就会看到……你可以在那儿直接用这个 skill。你也随时可以去插件市场（marketplace）看看。我其实不确定还有没有别的纯 skill 的插件，但 P-Stack 算一个。不过我们很多插件本身也自带 skill，也就是说除了 MCP server 之外还附带 skill。比如说，如果你——我想想——比如你装了 X 插件，我记得它除了工具之外还带了一个 skill。然后对，所以

[32:24] Lauren

plugins can contain like lots of different pieces of functionality that extend your grabbot instance um to do like lots of really interesting things.

插件可以包含很多不同的功能模块，扩展你的 Grok Bot 实例，让它能做很多特别有意思的事。

[32:35] Peter

Awesome. Uh, before we wrap up your share, like um I'm I'm curious about your social bots like um

太棒了。在你这部分分享收尾之前，我挺好奇你那些社交媒体 bot 的，比如说

[32:41] Lauren

yeah,

嗯，

[32:41] Peter

how how do you separate the signal from the slop

你是怎么从一堆垃圾内容里筛出真正有用的信号的

[32:44]

on X?

在 X 上？

[32:46] Lauren

Well, yeah. So, a lot of my social bots are connected to uh X as well as like LinkedIn uh which I guess I I tend to use X more than LinkedIn. Um but I have u uh I have like one bot called Potato of course who goes through all of the all of my mentions. So like uh I have this like meme on X where if you say potato three times I get like a notification. [laughter] Uh and that that that's this bot basically. this bot is like has a routine where it's checking x uh I think I set it to every 30 minutes or so uh to check for all the mentions uh and then I just jump in there and I'll often respond uh or it'll just kind of like aggregate like you know like you know five users have reported the same bug right and maybe you should take a look at that and so I I kind of think of the socials as not just not just managing my I guess my social uh networks but also as a form of like my outer loop where since I'm using the socials a bit a bit for work as well. Uh it means that you know I can stay on top of like the pulse of you know people are reporting bugs or they're running into issues or you know maybe uh a new feature that we launched is confusing you know like I want to be able to

嗯，对。我的很多社交 bot 都接了 X，也接了 LinkedIn，不过我用 X 比 LinkedIn 多。我有一个 bot，当然就叫 Poteto，它会把所有 @ 我的内容都过一遍。就是我在 X 上有个梗：谁要是连说三遍 poteto，我就会收到通知。（笑）其实就是这个 bot 在干这事。它有个定时任务（routine），会去 X 上查——我记得设的是大概每 30 分钟一次——查所有提到我的内容。然后我就直接进去，经常会回复；或者它会帮我汇总，比如：有五个用户报了同一个 bug，你也许该看一下。所以我觉得社交媒体对我来说，不只是用来打理我的……大概算是我的社交网络吧，它也是我外层循环（outer loop）的一种形式。因为我也会拿社交媒体做一些工作上的事，这意味着我能随时掌握动态：大家在报 bug、遇到了问题，或者我们新上线的某个功能让人看不懂，你知道，我希望能够

[34:03] Lauren

uh like stay connected to that and have that feed my inner loop of you know my engineers cuz then because these are all just bots right I can I can always say like I can go to like matcha and say like like hey uh you know uh potato said blah blah blah Uh, and then the context can just be grabbed there, right? I don't have to repeat all of the X context that I got. I don't have to copy paste anything. I just say go talk to potato and figure it out, right? And come back to me with a design proposal for how we fix this issue. Mh.

跟这些保持连接，并让它喂给我的内层循环（inner loop），也就是我的那些工程师。因为这些全都只是 bot，对吧，我随时可以……比如我可以去找 Matcha，说：嘿，Poteto 说了巴拉巴拉。然后上下文直接就能从那边抓过来，对吧？我不用把从 X 上拿到的上下文再复述一遍，也不用复制粘贴任何东西。我只要说：去找 Poteto 聊聊，把情况搞清楚，对吧？然后拿一份怎么修这个问题的设计方案回来给我。嗯。

[34:36] Lauren

So I think like yeah when you for me the the the the power of grapa comes when you connect all of these different things together and have them kind of uh I don't know if there's a better word for like but synergize you know like the the context can can be grabbed from lots of different places um and kind of be grouped together in a way that is much richer than if I were to type it all out by myself.

所以我觉得，对我来说，Grok Bot 真正厉害的地方，在于把这些各不相同的东西连到一起，让它们……我也不知道有没有更好的词，就叫「协同增效」吧。上下文可以从很多不同的地方抓取过来，再汇聚到一起，比我自己一个字一个字全敲出来要丰富得多。

[35:01] Peter

Yeah. Yeah. I mean like context is everything with this AI stuff and um Yeah.

对，对。我是说，搞 AI 这些东西，上下文就是一切。嗯，对。

[35:05] Peter

Yeah. Basically all all these bots live in the same house and they can talk to each other and like you know pass conscious to each other, right?

对。基本上就是，这些 bot 全都住在同一栋房子里，彼此可以交流，还能互相传递上下文，对吧？

[35:10] Lauren

Yeah.

对。

[35:11] Peter

Cool. Uh and uh where where does potato come from? Do you just really like potatoes? [laughter]

酷。那 poteto 这个名字是怎么来的？你就是特别喜欢土豆吗？（笑）

[35:18] Lauren

Uh yeah, it was just like a I don't even know. I think it was like when uh this was like many years ago and I was just playing like an M MMO and I was like I couldn't come up with a name and I just I I have a thing for food as you can tell from all my bot names.

呃，对，就是……我自己都说不清。我记得是很多年前，我在玩一款 MMO（大型多人在线游戏），怎么也想不出该起什么名字。而我对食物情有独钟，从我那些 bot 的名字你们也看得出来。

[35:32] Lauren

So I was like oh potato why not and I because I couldn't be I wasn't really thinking that hard and then potato with the normal spelling was taken. Mhm.

所以我就想，哦，potato（土豆），有何不可？反正我也没怎么认真想。结果正常拼法的 potato 已经被人占用了。嗯。

[35:41] Lauren

So I was like, "Oh, okay." I was I was actually learning Japanese at the time. So I was like, "Oh, okay. Maybe you spell it like potato, like in like the Japanese katakana." And so and that and that kind of stuck ever since.

我就想：「哦，好吧。」我那时候其实正在学日语，所以就想：「哦，好吧，那要不就照日语片假名的读法拼成 poteto。」然后这个名字就一直沿用到了现在。

[35:52]

Got it. Got it. Okay. Cool. All right. Well, this is super helpful. Yeah. And I I just want to ask a few wrap-up questions. So I I I feel like when I look at the Grockbot interface, it's almost like a version It's kind of like a version of Slack where you have your DMs, your teammates, right? And then you can add them to your group chat and stuff. And and you don't have to share any secrets, but like there's been a lot of stuff online about how like the stuff is still very single player AI. It's like one human talking to a bunch of bots. Maybe you can talk a little bit about like how you can make the stuff like so like you and so Lauren and Pen can also talk to a bunch of bots together or or something like is there is there any interest in doing that or ping you want to talk about it? Yeah, I think uh when we started designing Grabbot, we started from a mindset of um how can we enable people to graduate from manually manage the task themselves because personally I see it like three different tiers of working like in the past we work in a context ourselves. We draw rectangles in Figma, we typing text in ocean, we sending messages in Slack and now people already got into a stage that we are like prompting AI to do that for us. Um but in a lot of setup um that you've seen people are doing right now like I believe like we are moving towards in the future that um instead of like manually managing all those bots uh people are started designing the system design the guardrail um how do you make the bots work together um with other human um with you um autonomously that was that agency and intelligence ingested and a lot of AI agent products was very model ccentric uh because you can see from the UI it's very model oriented uh we built harness around the model and then we build product around those harness so the interface was like tailored for like along the way we created a lot of technical jargon and

Peter： 明白了，明白了。好，酷。好的，这些真的特别有帮助。那我想再问几个收尾的问题。我感觉看 Grok Bot 的界面，它几乎就像……有点像某种版本的 Slack：你有私聊、有队友，对吧？然后可以把他们拉进群聊之类的。你们不用透露任何机密，不过网上一直有很多讨论，说这类东西目前还是非常「单人模式」的 AI——就是一个人对着一堆 bot 说话。也许你们可以聊聊，怎么让它变成……比如 Lauren 和 Peng 也能一起跟一堆 bot 对话之类的？你们有没有兴趣做这个？或者 Peng，你想来聊聊吗？

Peng： 好。我觉得我们刚开始设计 Grok Bot 的时候，出发点就是：怎么让人们从「亲手管理任务」这个阶段毕业。因为我个人把工作方式分成三个层级。过去是我们自己亲手干活：在 Figma 里画矩形，在 Notion 里敲字，在 Slack 里发消息。现在大家已经进入了下一个阶段，就是写 prompt 让 AI 替我们做这些。但从你看到的很多人现在的配置来看，我相信我们正在走向这样的未来：人们不再手动管理所有这些 bot，而是开始设计系统、设计护栏（guardrail）——想清楚怎么让 bot 带着被注入的能动性和智能，跟其他人、跟你一起自主协作。而很多 AI agent 产品都非常以模型为中心，从 UI 上就看得出来，完全围着模型转：我们在模型外面搭 harness（包在模型外面、管提示词/工具调用/循环的那层程序），再在 harness 外面搭产品，所以界面就是为了……量身定做的。一路下来，我们造出了一大堆技术黑话和

[37:39]

concepts that like my mom would never know what is a soldout MD so um it's an experiment to like see how can we make the concept the system model match this user's mental model and then we found this like personified named agents worked pretty well and then it resonate with people because that's how you operate with the physical world today. Like you don't need to know how the car insurance company work but you know who to call like you you have an agent for that. If you have a car accident you don't need to know how the medical system is working but as soon as you have uh health insurance you have this doctor's office all those infrastructure in the real world. So have that point of contact become your like main input output interface. um is a paradigm shift for me because we see a like internal like workflow become I catch a b bunch of bugs then I just take screenshots and annotate them and then dump into a bot and then it will create and post it into this notion page that my engineers bots would pick it up from there. So it's kind of like we started to use the app in a very indirect way or agentic way. Uh but somehow it was able to like connect all the dots together and then organically grows into like new workflows in a very autonomous and long running way. Maybe that's like a general best practice overall using this stuff like it's just like because in some ways like the humans in a loop because we're also passing context from one to another and like kind of try to provide feedback. And I I don't know. I'm always a little bit skeptical when people say, you know, I'm just going to go to bed and let this bot do everything [laughter] and like, you know, just make a data design or build me app and like I'll come back and it's perfect. But I guess maybe you guys have more experience doing this. Like do you guys have any tips for maybe letting these bots run longer without going off the rails or

Peng： 概念，像我妈就永远不会知道 soldout MD 是个什么东西。所以这算是一次实验：看看怎么让产品概念、让系统模型去贴合用户的心智模型。然后我们发现，这种拟人化、有名有姓的 agent 效果挺好，也能引起大家的共鸣，因为这就是如今你在现实世界里做事的方式。比如你不需要知道车险公司是怎么运作的，但你知道该给谁打电话——有个代理人（agent）专门替你管这事。出了车祸，你也不需要知道医疗体系怎么运转，只要你有医保，就有诊所，有现实世界里那一整套基础设施。所以，让那个对接人成为你主要的输入输出界面，对我来说是一次范式转变。因为我们看到，内部的工作流变成了这样：我发现一堆 bug，就截图、标注，然后一股脑丢给一个 bot，它会整理好发到一个 Notion 页面上，我的工程师 bot 再从那里接手。所以我们开始用一种很间接、或者说很 agentic 的方式在用这个 app。但不知怎么的，它就是能把这些点都串起来，然后有机地长出新的工作流，而且非常自主、能长时间持续运行。

Peter： 也许这就是用这类东西的一条通用最佳实践吧。因为从某种程度上说，人还在循环里（human in the loop）：我们也在把上下文从一处传到另一处，也在试着给反馈。我也说不好。每次有人说什么「我去睡觉了，让这个 bot 把活全包了」（笑），比如「直接把设计做了」「给我搭个 app」，等回来一看就完美了——我总是有点怀疑。不过你们在这方面可能更有经验。你们有没有什么窍门，能让这些 bot 跑得更久又不跑偏，或者

[39:28] Peter

like kind of actually producing stuff that you want other than using Lauren scales? Like is is there other tips to to do this? Yeah.

真正产出你想要的东西——除了用 Lauren 的那些 skill 之外？还有没有别的窍门？嗯。

[39:34] Lauren

Uh I think the general thing for me is Yeah. I mean you don't have to use like PAC or anything. But I think there is something to do with with skills. Um, and I I talk about this a bunch, but I think it ultimately comes back to trust, right? Like how much trust are you willing to give your agent or your bot to be able to do a task for you?

嗯，我觉得对我来说，大方向是——对，你不一定非得用 P-Stack 什么的。但我觉得这确实跟 skill 有关系。这个我讲过很多次了，但我觉得归根结底还是信任，对吧？你愿意给你的 agent 或者 bot 多大的信任，让它能替你把一件事做成？

[39:55] Lauren

Yeah. And I think one thing I always try to encourage people to do, you know, whether you are a PA stack user or not is to actually spend the time to sort of observe the bot or agent doing the work and then you course correct it. But then at the end of the the conversation, you turn that into a skill, right? You you you turn that into a reusable skill. So, for example, I don't know, maybe you have um um a skill for or you you ask your bot to, you know, fill up an expense report form, right? Based on go through all of my, you know, expenses this month and file the, you know, put it put it all into the expense tool, find the receipts for my email, blah blah blah. Um, and once you've kind of iterated a lot on that first loop, right, if you, you know, you want to be able to get your skill good enough to the point where you can do like slash, you know, expense report, right? and it just does it well, right? Yeah.

对。我一直鼓励大家做的一件事——不管你是不是 P-Stack 的用户——就是真的花时间去观察 bot 或 agent 干活，然后随时纠偏。等这段对话结束时，再把它沉淀成一个 skill，对吧？把它变成一个可复用的 skill。比如说，我也不知道，也许你有个 skill 是……或者你让 bot 去填一张报销单，对吧？就是「把我这个月所有的开销过一遍，全部录进报销工具，从我的邮箱里找出收据」，等等等等。等你在第一轮循环上反复迭代了很多次之后，你会希望把这个 skill 打磨得足够好，好到你只要敲一个「斜杠 expense report」，它就能做得很好，对吧？嗯。

[40:48] Lauren

And when you've got gotten to a point where you can just do like a one shot, right? You just say slashexpense report and it just does the the thing for you,

等你到了能一次搞定（one-shot）的程度，对吧？你只要说一句「斜杠 expense report」，它就把事情给你办了，

[40:57] Lauren

then that actually gives you much more confidence to walk away and say, "Okay, now you can just go off and automatically, you know, uh set up a routine, right? Like every time I have a new email come in that has an receipt, you know, automatically file this report for me." Um, and you don't have to babysit it as much. But I think it yeah there's no secret to that in my opinion is like whether it's like doing design work or engineering work or life admin work it all comes back to how much trust have you get you do you have in your bot and if you don't trust it yet then you know there's things like skills uh you know prompting that kind of really help you build that trust uh but yeah I think you need to it has to be kind of gradual it's very hard to get someone to you know who's maybe never used a bot before or they don't have a lot of skills set up and then say, "Okay, now just, you know, you can walk away." Of course, they're not going to be able to trust it, right? Because they haven't seen it do, you know, what it can do. But once you've done it once or twice with the agent and you know that they can repeat that,

这时候你就会有足够的底气撒手走开，说：「好，现在你可以自己去跑了，自动地——设一个 routine（例行任务），对吧？比如每次有带收据的新邮件进来，就自动帮我把报销单提交了。」这样你就不用那么时时盯着它了。但在我看来，这事没什么秘诀：不管是做设计、做工程，还是处理生活杂务，归根结底都在于你对自己的 bot 有多少信任。如果你还不信任它，那就有 skill、prompt 这类东西，能真正帮你把这份信任建立起来。不过，对，我觉得这必须循序渐进。对一个可能从没用过 bot、或者没配过多少 skill 的人，你直接跟他说「好，现在你可以撒手不管了」，这太难了。他们当然没法信任它，对吧？因为他们还没见识过它能做到什么。但只要你跟 agent 一起做过一两次，知道它能重复做到，

[42:01] Lauren

then you can actually free yourself up and

你就真的可以把自己解放出来，然后

[42:04] Lauren

free, you know, go do other things, right? Because now you can trust your bot will do the right thing.

腾出手去做别的事，对吧？因为这时候你已经可以相信，你的 bot 会把事情做对。

[42:08] Peter

So, okay. So, okay. Okay, so I guess last question is if I just download Grockbot, then it sounds like the steps are maybe set up a bunch of bots for your manual workflows, go through them manually, and then if you can do a comfortable state, tell them to build a scale and then eventually set up some routines to like do it proactively. Is that kind of what you guys recommend?

好，好的。那我想最后一个问题是：如果我现在下载 Grok Bot，听起来步骤大概是：先针对你那些手动的工作流建一批 bot，手动带着它们走一遍；等你到了放心的状态，就让它们把流程做成一个 skill；最后再设置一些 routine，让它们主动去做。你们推荐的大概就是这样吧？

[42:26] Lauren

Uh that that's my flow. I'm kind of curious paying about your your own

嗯，这就是我的流程。我挺好奇 Peng 你自己的……

[42:30] Peng

Yeah, I definitely agree. It's very gradual process cuz um typically I would recommend like start simple just give it a task and like cuz until today I'm still testing the limit of it. Basically I'm trying everything I touch my keyboard and mouse and try to dedicate it to it and see how it works.

对，我完全同意，这是个非常循序渐进的过程。我一般会建议从简单的开始，就先给它一个任务。因为直到今天，我都还在试探它的极限。基本上，凡是我要动键盘鼠标去做的事，我都会试着交给它，看看效果怎么样。

[42:47] Peng

So you almost have to overstretch it to see where the limit is. Um and also I would describe this as like a building a relationship between you and the bot. um because it's very intentional design that all the capabilities are shared, the tools, the connectors um the skills um but the memory and preferences are local to each individual bot. So it's kind of like abstract those um reusable components into the role um to help people manage and then like to form this relationship. It takes a bit time but like try it with start a simple task then start a recurring task and then do this set up routines let it automate it do it and maybe it like burns too much token and tone down a bit and then get connected to more tools skills um customize it so it's like gradually it gets more mature and it knows you so it's even a very personal thing I mean like you knows yourself better and then your assistant definitely would be different from my assistant. So, it's this like mutual um progress.

所以你几乎得把它用到超出负荷，才能看清极限在哪。另外，我会把这形容成在你和 bot 之间建立一段关系。因为我们是刻意这样设计的：所有能力都是共享的——工具、连接器（connector）、skill——但记忆和偏好是每个 bot 各自独立保存的。这有点像把那些可复用的组件抽象到角色里，帮大家去管理，进而形成这种关系。这需要一点时间，不过可以先从一个简单任务开始，再试一个周期性任务，然后设置 routine，让它自动去做；也许它烧了太多 token，那就收一收；再给它接上更多工具和 skill，做些定制。这样它会慢慢成熟起来，也越来越了解你。所以这甚至是件非常私人的事——你最了解你自己，你的助理也肯定跟我的助理不一样。所以这是一种双向的共同成长。

[43:58] Peter

It's it's like onboarding employee, right? You can kind of trust them more over time.

这就像带新员工入职，对吧？时间久了，你就会越来越信任他们。

[44:02] Lauren

Yeah, I feel like that. I feel like that's a very good way to put it. It's like kind of hiring employ virtual employees, right? And then you want to be able to trust them. So, the same way like you would have to train a new employee before you, you know, let them be autonomous. I think if you set that up then you can actually get into a really nice flow where you as the sort of CEO right you can now delegate and step away and do like higher level strategic things [snorts]

对，我觉得是这样，这个说法特别好。有点像在招聘虚拟员工，对吧？然后你希望能信任他们。就像你得先培训新员工，才能放手让他们自主干活一样。我觉得只要把这套搭起来，你就真能进入一种很顺畅的状态：你差不多就是那个 CEO，对吧，现在可以把活委派出去、自己抽身出来，去做更高层面的战略性工作。（笑）

[44:28] Peter

now now we can all be the CEO of the product now

这下我们人人都能当产品的 CEO 了。

[44:30]

yeah exactly [laughter]

对，没错。（笑）

[44:32] Peter

cool well well thanks so much both for walking through your setup and everything else I guess Lauren people can find you on potato is that is that the handle

好，非常感谢两位带我们过了一遍你们的配置，还有其他种种。Lauren，大家可以在 poteto 上找到你，这就是你的账号对吧？

[44:40] Lauren

yeah uh you can find me on x uh potato with a e p o t e t o

对，大家可以在 X 上找到我，poteto，带 e 的那个：p-o-t-e-t-o。

[44:45] Peter

and pen do you want to stay anonymous or do you want to share your handle?

Peng 呢？你是想保持匿名，还是也分享一下你的账号？

[44:48] Peng

Oh, I'm also active on X. Uh my handle is my name P N G Z H N G and then underscore.

哦，我在 X 上也挺活跃的。我的账号就是我的名字 P-E-N-G-Z-H-E-N-G，后面再加一个下划线，也就是 pengzheng_。

[44:54] Peter

All right, guys. Well, thanks so much for your time and uh Grapa is an amazing product. Congrats on building it and and yeah, I'm excited to see what's next.

好了，各位，非常感谢你们抽出时间。Grok Bot 是个很了不起的产品，恭喜你们把它做了出来。我很期待你们接下来的新动作。

[45:00]

Yeah, thanks for having us.

嗯，谢谢你邀请我们。

[45:01]

Yeah, thanks for having us. It was really fun.

嗯，谢谢邀请，聊得很开心。