设计师拥抱氛围编程与自主工具

AI 流畅度最终会变成一种如影随形的东西。我不认为我们会永远去强调这门语言。但现在它是一门重要的语言，因为很多公司会说：‘是的，我们想要的设计师要能进来思考；我们希望有人知道如何应对“what if”这个问题，而且不能让人觉得那是 2012、2015，甚至 2018 年的思路。’这个问题似乎正由‘你展示什么、分享什么’来回答。社交媒体现在就是新的发现引擎。不管好坏，那就是证明你是一个在尝试、在实验的人的方式。

AI fluency is going to eventually just become like an ambient thing. I don't think we're going to like highlight that language forever. But I think right now it is an important language because a lot of companies are like, "Yeah, we want this designer who can come in and think. We want someone who knows what to do with the question 'what if' in a way that doesn't feel like it's from 2012, 2015, you know, 2018." That question seems to be being answered by like, "What are you showing that you're sharing?" Like social media is the new discovery engine right now. Like that is the way to — for better or worse — that is the way to prove that you are someone who is experimenting.

欢迎收听 Dive Club。我是 Reid，在这个节目里，设计师永远都在学习。本周的嘉宾是 Tommy Geoco，他刚刚走访了当今一些领先的设计团队，比如 Vercel、Perplexity、Ramp，我们将深入聊一聊他在 AI 工作流方面看到的种种，以及设计的下一站。首先，我想从他最近的《原型设计现状调查》入手，看看哪些细节最突出。

Welcome to Dive Club. My name is Reid and this is where designers never stop learning. This week's episode is with Tommy Geoco who just finished visiting some of today's leading design teams like Vercel, Perplexity, Ramp, and we're going to go deep into everything that he's seeing around AI workflows and where design is headed next. And to start, I wanted to dig into his recent state of prototyping survey to figure out what details stood out the most.

最突出的一点是，目前正在 vibe coding 的设计师对自己的工作流更满意。我听到很多‘不’，而且我们都听到了很多噪音，老兄。这我没想到。我非常好奇，因为我想：‘我知道这发生得很普遍。但有多少人这么做？他们满意吗？’因为我本来以为——我们听到过类似 Meta 用 token 层或排行榜来激励员工的故事，对吧。你知道，就是这样被硬塞给员工。但在这个调查里，大家似乎还挺喜欢的，我觉得这真的很特别。我还觉得——我希望能把数据解析得再细致一点——但那些目前占了大多数时间的人，比如如果他们的工作流大部分时间都花在 vibe coding 上，那是设计工程师，这说得通。主管级、负责人级，也行。然后是那些非设计师角色，可能是学生和研究人员；再是管理者；再是一般的中初级个人贡献者（IC）。让我觉得有意思的是，管理者的占比高于中初级 IC。要么是事情在往下渗，大家在尝试然后把经验传下去——我们在工作中看到的就是这样；但另一种可能是，有些管理者或团队还没有为团队其他成员腾出空间。我看到一大批中初级设计师正在被排除在流程之外，我的数据也说明了这一点。

The number one thing that stood out to me was that designers who are currently vibe coding are more satisfied with their workflows. I hear a lot of 'no,' and we both hear a lot of noise, man. And I did not expect that. I was very curious to see because I was like, "I know this is happening a lot. How much, and are people happy with it?" Because I assumed — I mean, we hear the stories like Meta incentivizing token layers or leaderboards, yeah. You know, and this like just like being crammed down the throats of employees. People seem to dig it in this survey and I thought that was really unique. I also thought — and I wish I had parsed this a little differently — but it's the people who are currently doing the majority, like so if a majority of their workflow is spent on vibe coding activities, it's design engineers, that makes sense. Lead principals, okay. But then it's like non-designer roles which might be students and researchers, then it's managers, and then it's like your general junior mid-level IC. And that part was fascinating that managers are doing more than junior and mid-level ICs. Either things are trickling down and people are experimenting and then they're going to pass it, you know, they're going to pass learnings down, which is kind of what we've seen on location, but it also might mean that like some managers or teams haven't yet made room for the rest of the team. I'm looking at a bunch of junior and mid designers that are getting cut out of the process as my data says.

是。是，而且，而且，对我来说，你知道，我想尽量往好处想，但归根结底，是的，公司需要腾出更多空间。我不知道趋势是不是这样，但我确实开始在某些类型的公司里越来越多地看到这种情况。

Yeah. Yeah, and, and, and that to me it's like, you know, I want to, I'm trying to give a little benefit of the doubt, but ultimately it's like, yeah, we need to make more room at companies. I don't know if it's trending in that way, but I'm starting to see more and more of that at certain types of companies anyway.

我觉得 vibe coding 的类型也很有意思。因为有的人会构建各种复杂的流程、系统或文件，让 Figma Make 甚至 B0 这类工具看起来、用起来都像他们的设计系统；而另一些人就是每天打开生产代码库，直接开始探索和上线。这之间是有很大的鸿沟的。而且有很多人来找我，他们很沮丧，因为他们想把这些工具拧成自己想要的形状。还有另一类设计师——我的意思是，如果你做的事情只是用代码把设计发布出来，最终进入产品本身，那设计工程师和设计师的区别是什么？这就是那条界线。我们现在甚至没有合适的语言来谈论这些不同的角色。

I think it's interesting to think about the types of vibe coding, too, because the experience of somebody who is building all of these crazy processes or systems or files to get a tool like Figma Make or even B0 to look and feel like their design system versus somebody who's just every day they open up the production codebase and just start exploring and shipping. That is a pretty big chasm, too. Where I get a lot of people coming to me that are frustrated because they're trying to bend these tools to their will, and then there's another class of designer who really is — I mean, what's a design engineer and what's a designer if all you do is ship designs in code that ultimately make it into the product itself? That's the line. Like we don't even have the right language to talk about these different types of roles now.

是啊。

Yeah.

然后他给我看了那个东西，我说：‘好吧，这确实够得上这个头衔。真的很酷。’我看到越来越多的人在尝试招聘品牌工程师。我以前甚至没想过这会是一个角色。但现在它确实存在。品牌工程师正在崛起。我觉得所有‘某某工程师’都会崛起。

And then he showed me that, and I'm like, "Okay, that kind of qualifies it actually. That's really cool." I'm seeing way more people trying to hire brand engineers. I never even considered that as a role recently. That is a thing. Brand engineers are on the rise. I think every blank engineer is going to be on the rise.

销售工程师、营销工程师。你知道，不管好坏，我觉得所有人都在向某种构建者压缩。

Sales engineer, marketing engineer. You know, I feel like everyone is compressing to, for better or worse, builders of some sort.

我昨天和 Snowflake 的设计工程师 Will King 聊过。他说：‘我懂。我热爱系统思维。我喜欢所有东西都在朝这个方向发展。你知道，我们在建工厂，我们在进一步工业化。但那种自由形态的创意工作流、那种探索的部分确实在流失，有些人必须去面对这个问题。问题就是：那部分会去哪里？它会被彻底消灭，还是在另一个地方重新长出来？是不是每个人都必须成为构建者？’我当时觉得这个问题很有意思。我自己很喜欢构建、设计系统和系统思维这一面。但我也承认，那部分是需要面对的。

I talked to Will King, design engineer at Snowflake, yesterday. He's like, "I get it. I love systems thinking. I like that it's all going towards this. You know, we're building factories. We're industrializing this even further. But there is like a loss of the free-form creative workflow, the exploration part that some people are just going to have to reckon with. And it's like, 'Where does that go? Does it just get eliminated, or does it sprout up in a different area? Like does everyone have to be a builder?' And I just thought it was an interesting question. I love the building and design systems and just the systems thinking side a lot. But I recognize that that's something to reckon with.

你花多少时间在那类事情上？

How much time do you spend on that stuff?

更准确地说，是它的性质变了，对吧？我其实经常思考这件事。我甚至有点对此不太好意思，但我的绝大多数——你知道的——杂乱的探索，现在都是用 AI 来做的。

It's more that the nature of it has changed, right? And I actually think about this a lot. Part of me is almost slightly self-conscious about it, but I do the vast majority of my — you know — messy explorations with AI now.

我感觉自己已经能够扮演所谓的“创意总监”角色了：我只是让 AI 给我看一个东西的 50 种不同做法，然后挑出我喜欢的部分，再重新组合，最终到了一个我会觉得“嗯，这样不错”的地方。然后我把这个从纸面上拿下来，用 Claude Code 跑一遍，它就出现在 localhost 上了。接着我再抠细节，在代码里做真正的精确设计。真是太疯狂了，老兄。这跟我职业生涯里任何阶段的工作流程都完全不同。有时候我会想：我是不是在失去那种手动操作、按 Command-D 一点点调整时的批判性思考？然而我又能探索一个比我一天能做的还要大得多的方案谱系，而且只用几分钟就能完成。我总会看到一些本没打算看到的东西。但这就是另一种不同，对吧？我只是换了一种贡献方式。我在锻炼不同的肌肉，可能也在让某些肌肉萎缩。谁知道呢。这种改变对我来说会产生什么影响，连我自己这个专业人士也还是个待定状态。

I feel like I've made the jump to the, quote, creative director, where I'm just working with AI to show me a certain thing 50 different ways. Then I pull the pieces I like, combine them again, and finally get to somewhere where I'm like, "Yep, that's good." Then I take that from paper, run it through Claude Code, and now it exists on localhost. Then I sweat the details and actually do the precision designing in code. Which is crazy, man. That's a very, very different workflow than anything I've done at any point in my career. Sometimes I wonder: am I losing some of the critical thinking associated with doing things manually—hitting Command-D and making tweaks myself? And yet I'm able to go through a spectrum of explorations so much larger than I could do in a day, and I can do it in minutes. I always see things I didn't plan on seeing. But that's different, right? It's just a different thing that I'm bringing to the table. These are different muscles I'm growing, and maybe different muscles I'm atrophying too, you know? I don't know. I'm kind of TBD on what the impact of that is for me even as a professional yet.

很快说一个事，然后我们回到节目里。最近人们从 ChatGPT 和 Perplexity 这类工具获取答案的频率，已经大大超过了点击 Google 的次数。所以，如果你的网站不能被 AI 读取，你就在你的分析后台里看不到某种可见性的流失。Framer 刚刚推出了一个免费的 AEO 工具，它会展示这些 AI 系统如何读取和理解你的网站。10 秒之内它就能扫描你的网站，给出一个 AEO 就绪度评分，并指出是什么在损害你的可见性。我在 Dive Club 网站上跑了一下，只得了 50 分（满分 100）。还好 Framer 会告诉我需要做的修复步骤。所以，请务必花一分钟用 Framer 的 AEO 扫描器检查你的网站。你只要访问 framer.com/aeo 就能开始。

Just a quick message, then we'll jump back in. People are getting answers from tools like ChatGPT and Perplexity a lot more than they're clicking on Google lately. So if your site isn't AI-readable, you're losing visibility you can't see in your analytics. Framer just launched a free AEO tool that shows you how AI systems read and understand your website. In ten seconds it can scan your site, score it on AEO readiness, and surface what's hurting your visibility. I ran it on the Dive Club website and only got a 50 out of 100. But thankfully, Framer gives me the exact steps to fix it. Definitely take a minute to use Framer's new AEO scanner on your website. Head to framer.com/aeo to get started.

好，听我说这个。Jitter 的最新版本让你可以在 Jitter 内部构建自己的创意工具，制作自定义动画和特效。你只要描述任何你想要的东西，Jitter 就会给你一套自定义参数，你可以用这些参数调整每一个细节——无论是 3D 控制、抖动，还是老式 VHS 滤镜。一旦把效果调到满意，你就可以保存、跟团队分享，并在不同项目里轻松复用。它就是简单到只要把你的想法打出来，非常适合像我这种没有丰富动画背景的人。有时候我还会让 Claude 帮我写提示词。这个功能真的把我的创意天花板炸开了，我已经等不及想看大家会做出什么了。去 dive.club/jitter 就能开始。好，我们回到节目。

Okay, get this. Jitter's latest release lets you build your own creative tools inside Jitter to make custom animations and effects. Just describe anything you want, and Jitter will give you custom parameters you can use to tune every single detail—whether it's 3D control, dithering, or even an old VHS filter. Once you get the effect where you want it, you can save it, share it with your team, and easily reuse it across projects. It really is as simple as typing out your ideas, which is perfect for someone like me who doesn't have a rich background in animation. Sometimes I even have Claude help me with the prompts. This feature has blown the roof off my creativity, and I can't wait to see what you make. Go to dive.club/jitter to get started. Now, back to the episode.

两个星期前我开始做一个事情：强迫自己每周有一天完全不用 AI。

I started something two weeks ago where I'm forcing myself to have one day of no AI.

不是吧。

No way.

因为我在构建这些 open-claw 系统，所以我知道这个感受是真实的：能把我的意图这类上下文放进去，感觉非常好——而且到处都是输入。我在想什么？我在做什么？我投了什么票？这段对话里有什么是值得我在 Twitter 上关注的？所以我一直在想各种办法，把真正有意义的 Tommy 世界观上下文弄进系统里。但后来我就会想：好吧，这确实很好，因为这个东西给我的输出非常丰富，于是我就开始把它当拐杖。所以不用 AI 的那天能让我真正重新把事情想清楚；而且那天结束之后，我会拥有更多真正由我自己写出来的上下文，再把它喂回去。所以它有两个目的。我也看到你确实能感觉到这一点——就像你，在没它的时候，生产力感觉真的很不一样。一开始我真的会觉得很受挫。嗯，我觉得对我来说最大的区别是：我不太再从一张空白画布开始思考了。我的思考已经足够给 AI 一套不错的护栏，让它往我想去的方向探索。但有时候，如果我要设计一个新布局、新组件或者网站的一部分，我会直接运行一条提示词，让 AI 在画布上做一堆事；在看见结果之前，我不会去批判性地思考那个设计。我会看到 20 个方案，然后才会想：“哦，那个有意思”或者“哦，也许我们应该用这个”。但我的批判性思考是从那个时刻——也就是我已经在看着某个东西的时候——才开始的，而不是从一张空白画布开始。也许我会打字、画箭头、画一些潦草的小草图。那些东西现在在我的流程里没有位置。它就只是不同了。这是一道从概率式设计到确定性设计的光谱，而我很喜欢在这一端工作。

Where I know it's a thing, because I've been building these open-claw systems, and it's really nice to get the context inside of that—like my intent, with inputs everywhere. What am I thinking? What am I working on? What did I vote? What was worth my attention on Twitter in this conversation? So I'm trying to figure out all these ways to get really meaningful Tommy's worldview context into the system. But then I'm like, okay, it's really nice because my outputs are really rich from this thing, and then I start to become a crutch. So no AI allows me to really think through things again. And also, I leave the day with more context that was truly authored by me, and I can feed it back. So it kind of has two purposes. And I'm seeing how you definitely feel it—like you, when you don't have it, the productivity feels really different. I feel stunted at first, for sure. Yeah, I think the biggest difference for me is I don't do near as much thinking from a blank canvas. My thinking is enough to give AI a pretty good set of guardrails for it to explore in the directions I want it to go. But sometimes, if I'm going to design a new layout, component, or section of a website, I'll just run a prompt to get AI to do a bunch of things on the canvas, and I won't think critically about that design until I see something. I'll see 20 things and then be like, oh, that's interesting, or, oh, maybe we should do this instead. But that's when my critical thinking starts—at the point where I'm already looking at something, rather than at a blank canvas. Maybe I'm typing words, drawing arrows, doing rough little sketches. That doesn't currently have a role in my process. It's just different. You have this spectrum from probabilistic to deterministic design, and I really like working over here.

我有个问题想问你，因为我真的很感兴趣。我们在线看到太多噪音了，而一个很好的做法就是直接问别人：“你现在的实际工作流程到底是什么样的？”你最近分享过一件事——你做了一些按钮，但那些按钮感觉像是 AI 做的，而且还完全符合 InFlight 的设计系统。那么你当时的过程是怎样的？它收集了哪些输入，直到你完全不在流程里，它就自动把这些按钮生成了出来？

Here's a question I have for you, because I'm really interested. We see so much noise online, and what's been really nice is to just ask people: what's your actual workflow right now? And you shared something recently—you had made these buttons, but it was like the AI had made them, and they were in the perfect design system of InFlight. So what was the process for you? What inputs were collected, to the point where you were completely out of the process and it just spun those up?

这其实要追溯到几个月前我开的一个 Claude 会话。和大多数 Claude 会话一样，我一开始说的是：我完全不知道要做什么。我觉得我应该用 Markdown 文件，更有效地把上下文传进代码库；我觉得你应该能根据提示词更有效地读取规则集、收集上下文。而我当时什么都没做，于是我就想：我该怎么办？然后我真的出去找了一堆 Dan Shipper 的文章、各种乱七八糟的东西，还有我收藏过的推文，全都喂给了它。我说：把这些消化掉，帮我思考一个方案，我完全不知道该怎么办。然后它生成了这样一张嵌套的上下文图——也就是贯穿代码库各层级的 Claude Markdown 文件——用来教它自己怎么完成某些事情。其中有一组视觉规则，描述了我喜欢组件以什么方式被构建。说实话，我从来没打开过那个文件，我其实不知道里面有什么。我只是定期地往里面贡献内容。比如，如果我得到一个不喜欢的输出，我就会说：“喂，不要再这样做了，我要的是这个。”然后它就自己去处理，把它改到某个地方，我再也不会去看它。

It actually ties back to a single Claude session I had maybe a few months ago. Like most Claude sessions, I started off saying, I have no idea what to do here. I feel like I should be using Markdown files and passing context through my codebase more effectively. I feel like you should be reading rule sets and gathering context more effectively given a prompt. And I wasn't doing anything, and I was like, what do I do? So I actually went out and found a bunch of Dan Shipper articles, a bunch of everything, and tweets I had bookmarked, and I just fed a bunch of stuff in here. I was like, consume this, help me think about a plan. I have no idea what to do. And it created this chart of nested context—Claude Markdown files throughout the codebase at different levels—that would teach it how to achieve certain things. And one of those is a set of visual rules for how I like components to be built. To be honest, I've never even opened that file. I don't actually know what's in there. I just regularly contribute to it. Like if I get an output I don't like, I'm like, yo, never do that again. This is what I want instead. And then it just goes and does something and goes somewhere, and I never look at it.

但这事我已经做得够多了，现在 Claude 会明白：如果我要做一个微妙的交互元素，Fred 可能希望我用这种三级样式，配这个悬停叠加混合，在渐变上稍微加一点透明度。现在这些我开箱即得。所有那些小内阴影，我也开箱即得。我再也不会得到带边框的东西了。我只会得到用内阴影来模拟边框的东西。而这只需要一次相当专注的会话。比如，我用 Claude 在这组文件上大概工作了 2 小时，不断打磨，把上下文喂进系统。然后从那里开始，就只是花点时间告诉它某个东西很烂，解释为什么，然后说“再也不要这样了”。这种事我已经做了大概 100 次。我真的觉得那种黑盒——因为我自己也完全一样——我很久没怎么看自己的代码了，我的开放 Claude 设置就像是我通过输出就知道它在匹配意图。我不会去读 Markdown 文件。谁有时间呢？

But I've done that enough now where Claude will know, okay: if I'm going to build a subtle interactive element, Fred probably wants me to use this tertiary style with this hover overlay blend that adds a little transparency on top of the gradient. And I just get that out of the box now. I get all my little inner shadows out of the box now. I never get things with borders anymore. I only get things using inner shadows to simulate borders. And it just took one pretty dedicated session. Like, I worked with Claude for probably two hours on this set of files, just refining it and feeding context into the system. And then from there, it's just taking a beat to tell it that something sucked, explaining why, and then saying, 'never let this happen again.' And I've just done that like 100 times. I really think that sort of black box — because I do the exact same thing — like I have not looked at much of my code in a long time, and my open Claude setup is very much like I know it's matching the intent based on the outputs. I'm not going to read a markdown file. Who has time for that?

但我的意思是，这是一个非常快的、令人满足的循环。所以当我听到人们说 GPT 5.5 不如 Opus 4.7 时，我想说，那是什么意思？你在比较什么？你只是说你一次性输入了一个东西，然后你在那一天从其中一个而不是另一个得到的那个概率性输出更好还是更差？因为那不是一种很好的评估方式。就像，它开起来怎么样？你知道，有种感觉是，你得开它们一阵子。当你带着意图去开，并且你只是不停地说“这里错了，修一下，那里错了，修一下”，然后迭代几次，我觉得你就能对它的驾驶感受有个好感觉。所以我从来不懂那种一次性的心态，比如“我比较了三个相同的提示词”。那就像你只是发射了一堆概率性的东西。我讨厌那些推文，兄弟。我讨厌那些推文。那甚至不算设计。我理解这对非设计师来说可能真的很有吸引力，但事实是，专业设计师的讨论中有太多都是关于哪个模型开箱就能产出更漂亮的东西，或者哪个模型的紫色渐变更少——我们到底在干嘛？AI 给了我职业生涯中从未有过的精度，然而我们总是很快说“好吧，AI 真行”。什么？这不是 AI 行。我在做一些一年前我还不知道怎么做微决策，而这只是因为 AI 给我的能力。现在真的很难谈这个。你必须真的跟人聊：你是怎么用的？你是怎么评估的？“哦，你走进了一个死胡同的兔子洞。”好，我们先把你拉出兔子洞，确保我们明白：这是你怎么开。好了，现在明白了“开”是什么意思，我们再聊一些次要的、能提供信息的东西。因为我觉得，一种正确的讨论方式正在出现——我知道设计师讨厌谈论使用工具，但这就是我们当下的处境。你怎么用这些东西？你怎么安装它们？所以对我来说，我现在怎么谈智能体？我觉得这是一种无头设计概念。我觉得你刚才描述的一切——你怎么创建那个按钮、这些文档、这个意图——我不认为未来会是这样，我们从一个工具到另一个工具，一遍遍重新发明。所有花在把我们的意图以正确方式拧进去的时间，会变成一个你随身携带的公文包，从一个工具带到另一个工具。你知道，就像 Ramp 的 Glass。他们意识到公司需要它，所以就在内部造了出来。他们就这样发布了。你知道，一个没有这些基础认知的人——它已经在他电脑上了，然后它就真的引导他上手。“好了，你现在连上 Workday 了。你在一个安全环境里连上了你需要的任何数据源。现在你有你的上下文公文包了。你想在哪用就在哪用。”我觉得在很多场景里都会是这样，比如设计。设计领域也会如此。我觉得其实 InFlight 会是一个很好的用例，如果你们决定往那个方向走的话。我当时就好像，你现在就在我的 iMessage 对话里。我还想到，家庭最终也会有这个，对吧？比如，我有一阵子跟你聊过我有个想法。你知道，你可以做一个叫 Hearth 之类的产品。它就是家庭的上下文层。就像你的 Life360——给不了解的人解释一下，它会在手机上追踪你的孩子，因为作为父母，你想确保他们说的是真的。或者它追踪他们开多快，因为如果你有个孩子在开车，你肯定想确保他们不超速。你可能有那些数据。你可能有家庭日历。你可能有所有这些。所以现在家庭成员不需要靠直觉去写自己的智能体式配置，而是有一个工具，有一个上下文层。当所有这些工具变得对消费者友好时，他们只要连上去，它就有上下文。我觉得这会出现在每一个垂直领域。我敢肯定。

But I mean, it's like a really fast, satisfying loop. And so when I hear people say GPT 5.5 is worse than Opus 4.7, I'm like, what does that mean? What are you comparing? Are you just saying you one-shotted a thing, and that probabilistic output you got from one versus the other on that given day was better or worse? Because that's not a very good way to evaluate it. How does it drive? There's this sense that you have to drive them for a little while. When you drive it with intent, and you're just saying, 'that's wrong, fix it, that's wrong, fix it,' and you do that a few iterations, I think then you have a good feel for how it drives. So I've never understood this one-shot mentality of 'I compared three of the same prompts.' It's like you just shot a bunch of probabilistic stuff. I hate those tweets, man. I hate those tweets. That's not even design. I get how that would be really appealing for somebody who's not a designer, but the fact that so much of the discourse from professional designers exists around which models put things out of the box in a more pretty way, or which models have less purple gradients — what are we doing here? AI gives me a level of precision that I've never had in my career, and yet we always are quick to say, 'well, AI slaps.' What? This isn't AI slapping. I'm making micro decisions that I didn't have a grid for a year ago, and it's only because of the abilities that AI gives me. It's really hard to talk about right now. You have to really talk to somebody about how are you using it? How are you evaluating it? 'Oh, you've gone down a rabbit hole that's a dead end.' Okay, let's pull you out of the rabbit hole first and make sure we understand: here's how you drive. All right, now that we understand what driving means, let's talk about some of the other things that are tertiary to it, that can inform. Because there is starting to emerge, I think, a right way to talk about — I know designers hate talking about using tools, but that's the moment we're in. How do you use these things? How do you install them? And so for me, it's like, how do I talk about agents right now? It's this headless design concept, I think. I think everything that you just described — how you created that button, these documents, this intent — I don't think there's a future where we're going from tool to tool and reinventing that over and over. All the hours spent just screwing in our intent the correct way is going to become just a briefcase you carry from one tool to the other. You know, like Ramp's Glass. They realized they needed it for their company, so they just built it internally. They shipped it this way. You know, someone who doesn't have this baseline education on how any of these work — it's already on their computer, and it just literally onboards them. 'Cool, you're now connected to Workday. You're connected to whatever data sources you need in a secure environment. Now you have your briefcase of context. Use it wherever you need to.' I think it's going to be the same in so many contexts, like design. That's going to be true for design. I feel like actually InFlight would be a great use case for this, if you guys decide to go that way. I was like, you're in my iMessage conversation right now. I think households eventually will have this, right? Like, I was talking to you at one point about this idea I had. You know, you call a product 'Hearth' or something. It's just the family's context layer. It's like your Life360 — for anyone who doesn't know, it tracks your children on their phones, because as parents you want to make sure they are where they say they are. Or it tracks how fast they're driving, because if you have a kid who's driving, you definitely want to make sure they're staying on the speed limit. You might have that data. You might have the family calendar. You might have all these things. And so now a family member doesn't have to vibe-code their own agentic setup, but they have a tool, they have this context layer. When all these tools become consumer-friendly, they can just connect to it, and it has context. I think that's going to be in every vertical. I've got to imagine.

是的，我想了很多，因为我现在有视觉语言的上下文。然后还有来自代码架构的上下文——里面有一些业务逻辑。我也有一些来自 Claude 的持续的头脑风暴上下文。比如，Claude 知道 InFlight 是什么，但总是不太容易接进去把一切都拉出来，或者可能有点过时，或者它在谈论一个我六个月前觉得酷但现在不怎么再想的想法。我对这个想法真的很感兴趣——你把 InFlight 拿出来说，但天哪，每当我和 Claude 即兴讨论时，我应该能把我收到的所有反馈作为上下文给它。它应该知道我的联合创始人 Kyle 总是吐槽我的设计缺乏留白。他总是觉得太密了。你知道吗？这些应该存在。或者如果我在设计某个界面区域，所有东西都会经过 Claude 或者，你知道，Codex——你选择的工具——这不重要，对吧？我喜欢你的公文包比喻，但那个模型应该能够无缝接入一年前关于那个产品区域所做的所有决策。没错。而且不应该需要你复制粘贴。

Yeah, I think about that a lot, because I have visual language context right now. And then there's context from the code architecture — there's some business logic in there. I also have some running brainstorm context from Claude. Like, Claude understands what InFlight is, but it's not always amazing to hook in and pull everything in, or maybe it's a little out of date, or it's talking about this idea that I thought was cool six months ago that I don't really think about much anymore. I'm really interested in the idea of — you tee up InFlight, but gosh, whenever I'm riffing with Claude, I should be able to give it all of the feedback I've received as context. It should know that my co-founder Kyle always hates on the lack of negative space in my designs. He always thinks it's too dense. You know? That should exist. Or it should, if I'm designing a certain surface area, everything's going to be going through Claude or, you know, Codex, your tool of choice — it shouldn't matter, right? I love your briefcase analogy, but that model should be able to seamlessly tap into all of the decisions that were made about that area of the product a year ago. Right. And there should be no copying and pasting from you.

它应该就能获取一切，你懂吗？就像设计大脑。YC 那条关于大脑的推文昨天其实有点火。我对它在大规模上的可行性持复杂看法，但如果你看某个单一垂直领域——也许是一个家庭，也许是设计场景——那会是我们工作方式中很大的一部分。

It should just be able to get everything, you know? Like the design brain. The brain YC tweet kind of went viral yesterday, actually. I have mixed opinions on the viability of that at a large scale, but when you're looking at a single vertical—maybe it's a family, maybe it's design context—that's going to be a huge part of how we work.

对，我觉得通用工具最终会有它的位置。我不认为现在就是正确答案。也许我不会直接断言它不是正确答案，但它引不起我的兴趣。我在 Ramp 问过 Diego 一件事：你们为什么做 Glass？为什么不用市面上那些通用工具，比如 Co-work 之类的？他说：因为 Co-work 擅长通用，而 Glass 擅长 Ramp。我觉得这种上下文恰恰是非常重要的一环，很多人还没真正体会过这意味着什么。所以我有了自己的 Claude，它非常懂 Tommy。不是说它会用 Tommy 的口吻说话，而是它对所有与 Tommy 相关的东西都很强。当你拥有这种东西，它和直接用 ChatGPT 或 Claude Code 做事情完全不一样。就像你说的，如果它能说：『嘿，我们当时是这么做的，但 Kyle 不会喜欢这个；你上周有个客户电话，他们提出了一个挺有意思的点，也许该把那标记成一个边界情况。』这真的很惊人。

Yeah, I think the general purpose tool might have a place eventually. I don't think that's the right answer right now. Maybe I won't just outright say it's not the right answer, but it's not interesting to me. One of the things I asked Diego at Ramp was, why did you guys build Glass? Why didn't you just use one of the general purpose tools that are out there, like Co-work or something? And he said, because Co-work is good at general purpose, Glass is good at Ramp. And it's like that context is just a really important piece that a lot of people haven't experienced what that means. So it's like I have my own Claude and it's really good at Tommy. It's not like, oh, it talks in Tommy's voice. It's like it's really good at everything Tommy. And when you have that, it's so different from just using ChatGPT or Claude Code for something. It's really, like you said, if it could be like, 'Hey, we did it this way, but Kyle's not going to love this, and you had a customer call last week where they called out something kind of interesting. Might want to flag that as an edge case.' That's pretty wild.

另一件事是，你知道，我们聊到这个的时候，人们立刻就会走极端。要么是手工的人类产出，要么是全面自动化。我不认为事情必须如此，而且我不喜欢全面自动化。有些人会说：『哦，我的智能体跑起来了，流程整夜都在跑。』那是我的噩梦。别这么干。如果是一个非常具体、边界清晰、需要反复迭代的微任务，我还能理解那样做的意义。但我想要保留摩擦。我发现一定程度的保留摩擦对我挺重要，这样我可以核对事实、检查意图，再重新调整方向。你知道吗，就像你开车的时候我睡着了，你开错车道了，我想让你回来。我希望自己能有机会更多这么做。而且那些车道，这个过程真的很有趣。至少对我个人来说，我并不觉得自己在失去自我或者被替代。我觉得自己更多是在掌舵和驾驶。另外，关于那个公文包的比喻：它就在那里，随叫随到，你手头可能有客户洞察，不需要主动去找。但谁说你不能就打开公文包，进去翻一翻、核对一些事实、把手弄脏呢？这只是一个渐进式披露的问题，对吧？而且我觉得人们有这种二元思维，认为 AI 就等于完全自动化。但其实不是——事实上有一些更好的工作流恰恰是那些没有完全自动化的。

The other thing too is, you know, we talk about this and immediately people are so binary about it. It's either a handcrafted human effort or fully automated. I don't think that has to be the case, and I don't like when things are fully automated. There are people who talk about, 'Oh, I got my agents, and I got these processes to run all night long.' That's my nightmare. Don't do that. I could see that being interesting if you have a really specific, well-outlined micro task that needs to be iterated on. But I want the friction. There's a preserved level of friction that I'm finding is kind of important to me, so that I make sure—let me fact-check intent, let me re-steer. You know, it's like you drove and I fell asleep, you drove in the wrong lane; let's get you back over here. I want to make sure I have the opportunity to do that more. And those lanes—it's a really fun process. At least for me personally, I don't feel like I'm losing myself or replacing myself. I feel like I'm very much steering and driving. And then the other piece with that briefcase analogy is that while it's there on demand, you have customer insights potentially, and you don't have to go in and find them. But who's to say you don't just open the briefcase and go, if you want, go fact-check some stuff, get in there and get your hands dirty? There's just a progressive disclosure problem, right? And I feel like there's this binary thinking that AI means fully automated. And it's like, 'No, in fact, I think some of the better workflows are the ones that aren't fully automated.'

对，同意。那你现在用的是——比如，你用什么——你的技术栈是什么？说吧。

Yeah, agreed. Are you using—like, what are you—what's your stack? Go ahead.

这不公平。我刚想问你来着。咱们快速交换一下——来吧，快速交换。

That's not fair. I was going to ask you. Let's—we'll trade really quick. We'll trade really quick.

但我会换个方式来问。我想让你设想一下：一个还没采用 AI 的传统组织一直在看你的 YouTube 视频，然后他们说：『这人太前沿了。』你确实是的。他们把你请进去，你基本上就像个 A 轮顾问。然后他们说：『嘿，Tommy，我们想在工具上省钱，所以准备签年度合同。作为一个设计组织，我们应该签哪些年度合同？』

But I'm going to ask the question slightly differently. And I want you to pretend that a legacy org that hasn't adopted AI has been watching your YouTube videos, and they're like, 'All right, this person's super plugged in,' which you are. They bring you in—you're like a Series A consultant effectively—and they're like, 'Hey Tommy, we want to save money on tooling, so we are getting annual contracts. What are the annual contracts that we should commit to as a design org?'

哦，天啊。对，这很难。这确实是个难题，因为我拥有的一个奢侈条件是：我没有很多企业公司都有的安全问题和合规问题。我们假设连 SOC 2 都不需要——我觉得很多这类工具如果没有 SOC 2，也马上会具备。所以假设是中等安全要求吧。那你真的可以按自己的判断，去选择那些能赋能组织、让组织变成你满意的样子的工具。好，对设计工作而言，这就是我看到的，也是我会推荐的。

Oh. Oh, man. Yeah. That's hard. That is a hard question because the one luxury I have is I don't have the security issues, the security and compliance problems, that I think a lot of these enterprise companies have. Let's imagine that SOC 2 even isn't a requirement—which I think a lot of these tools, if they don't have SOC 2, they've got to be right on the horizon, right? So let's imagine it's mid-level security requirements. You really can lean into what you think tools empower an org and shape an org in a way that you'd be happy with. Right. Well, for the work, for design work, this is what I'm seeing and what I would recommend.

首先从高层看，如果按原语来想，你有画布。我觉得画布这类东西在这个工作流里永远有一席之地。你的画布工具是什么？我觉得你需要某种代码意图工具。我非常喜欢 Codex。我真的很喜欢 Cursor Codex 那种面向智能体的新 UI 模式。这个流程里你需要有一个智能体式编码工具。然后，我觉得你需要你的公文包、你的上下文、那个把一切汇总起来的东西，它就像你所用两类工具的后端。然后你需要工作所在的表面区域，可能就是 Linear 那样的东西，对吧？也许你是 Atlassian 用户，他们在那里也有一个智能体工具——我忘了叫什么。所以你需要的是工作所在的表面区域，也就是工作唯一事实来源所在的地方。对我们来说，一支视频完成时，唯一事实来源并不是我们在 Notion 里写的脚本，而是那个东西被上传到我们后端的地方，也许就是 Frame.io，对吧？只要它进了 Frame.io，我们就知道它做完了。在那个地方，一旦上传，它就可以变成很多不同的产物：可以变成一个上传到 YouTube 的任务，也可以变成我们需要写的描述。所以你需要有一个地方，让团队知道工作在进行、工作意图是什么、以及工作在哪里完成。所以我认为像 Linear 这样的工具就是那个地方。

First from a high level, if you think about primitives, you have canvas. I think canvases are always going to have a place in this workflow. What's your canvas tool? I think you need some sort of code intention tool. I like Codex a lot. I really like the Cursor Codex new UI pattern for agents. You need to have an agentic coding tool inside of this process. And then I think you need your briefcase, your context, the thing that ingests everything, like a back end to the two types of tools you're using. And then you need the surface area where the work lives. And maybe that's like Linear, right? Maybe you're an Atlassian person and they have an agentic tool there—I can't remember what it's called. So you need the work surface area where the source of truth of work lives. For us, when a video is done, the source of truth is not the script we wrote in Notion. The source of truth is where that thing gets uploaded in our back end, which might be like Frame.io, right? If it's in Frame.io, we know it's done. That's a place where, if it gets uploaded there, it could become a lot of different artifacts. It could become a task to upload on YouTube. It could become a description we need to write. So you need somewhere where you guys know the work is happening, the intention of the work, and where the work is done. So I think a tool like Linear is that.

然后你需要界面，你需要 Slack。我真的希望 Slack 在关注这件事，因为通过 Slack 和智能体聊天，我觉得有很多地方可以改进。和智能体聊天非常不一样——我会在一个智能体下面滚动七个不同的聊天实例，这有点乱，而且很容易丢掉上下文。所以这需要更好的工具。但目前来看，Slack 仍然相当不错。我看到人们在 Slack 里的用法也还行。所以你需要一个界面层；你需要工作所在的那一层；你需要画布来处理更确定性的工作；你需要智能体，也就是你处理代码这类更概率性工作的界面。

And then you need the interface. You need Slack. I really hope Slack is paying attention to this, because talking with agents through Slack has a lot of areas where I think the process could be better. Talking with agents is very different—I'll scroll through like seven different chat instances with my one agent, and it's just kind of messy, and you can lose context easy. So there needs to be a better tool for that. But right now, I think Slack is still pretty good. The way I've seen people using it in Slack is good. So you need an interface surface area. You need where the work lives surface area. You need a canvas for more deterministic work. And you need your agent, or your interface for coding, for more probabilistic work.

你知道，说到画布类工具，目前我关注三个。这话可能会给我惹麻烦。我依然认为 Figma 还在局内。很多智能体式团队，哪怕是那些非常前瞻的团队，仍然大量使用 Figma。它依然是整个图景中很重要的一部分。

You know, when I think about a canvas, there are three that are on my radar right now. This will get me in trouble. I still think Figma is in the game. A lot of even a lot of these agentic teams that are super forward-thinking are still using Figma a lot. It's still a big part of the equation.

完全同意。只要 Figma 能继续创新，并兼顾智能体式这一面，我就会把它算在局内。Paper 没有进我们这次调查的前 30 工具，这让我很意外，因为我每次和人聊天都会听到：‘哦，我们最近在玩 Paper。’ 现在在画布类工具里，Paper 出现的频率比任何其他新工具都高。另外，pencil.dev 也很有意思。我觉得 Tom 是个闷声赢家。他会继续把那个产品做得非常好。还有一点是，当你把这两个工具放在一起想时，这不只是技术栈的问题——你需要转变对工作的思考方式。这正是最难的部分。你想想，当我们开始装流水线的时候，早期进来的人只是尝试在旧工厂流程上加新设备，结果行不通，对吧？直到福特站出来说：‘不，我们必须重新布置整个工厂。’现在也必须发生一次类似的转变。Metalab 有一个叫 Team Zero 的团队，大概三个人。他们的任务是：‘嘿，你选项目、选客户，只要客户同意。你要用全公司其他人不会用的方式去探索 AI，然后把结果汇报给我们。’他们有一套自己的标准来判断是否成功。我认为这是公司规避风险很聪明的办法，而不是指望一下子重塑整个工厂车间。这样一来，团队可以回来说，哪些有效、哪些无效，然后他们可以慢慢推进。而另一边，Ram 那些人就只是说：‘不，我们要让所有人都这么做。’这就是我看到的两种工厂车间重构方式。

Totally. And as long as Figma can continue to innovate, with the agentic side considered, I would consider them in. Paper didn't show up in our top 30 tools on this survey, which was wild to me, because every conversation I'm having, I'm hearing, 'Oh, and we're playing with Paper right now.' I'm hearing Paper pop up more than any of the other new tools right now on the canvas side. And then pencil.dev is really interesting. And I think Tom is a sneaky little winner. I think he's going to continue to make that product really good. The other thing is, it's not just when you think about stacks like these two—there needs to be a shift in how you think about work. And that's the hard part. You know, I think when we started installing assembly lines and stuff, people came in early and just tried to install new equipment on top of old factory workflows, and it didn't work, right? Until Ford came out and said, 'No, we have to rearrange the entire factory.' A version of that has to happen right now. Metalab has a team called Team Zero. It's like three people. They're tasked with, 'Hey, you pick the project, the client, assuming the client's cool with it. You're going to explore AI in a way that the rest of the company isn't, and you're going to report back to us how it goes.' They have their own set of criteria for how they determine success on that. And I think that's a really smart way for a company to mitigate the risk of hoping that they reinvent the whole factory floor. This way, this team can come back and just say, here's what's working, here's what's not, and they can slow-roll into it. Versus then you have the Rams who are just like, 'No, we're just like everybody do it.' And those are the two different types of ways that I'm seeing factory floor restructuring.

好，我们就以此为切入点吧。因为你刚刚实地走访了一些世界上最高产的设计团队：Perplexity、Vercel、Ram、Metalab。那么，跟我聊聊，你觉得这些顶级设计团队今天是如何运作的、他们在探索什么，有哪些共通的脉络或信号特别突出？

Okay, let's use this as a launching point because you've now gone on the ground with some of the most prolific design teams in the world: Perplexity, Vercel, Ram, Metalab. So talk to me a little bit about the unifying threads or signals that are standing out to you about how the best design teams are operating today and what they're exploring.

我认为最关键的一点是，公司要为试验留出空间，而这很昂贵。它在 token 成本上很贵，而且你会走进很多兔子洞，大量工作最后都被丢弃。领导者会说：‘不，我要把它解封。我们会投入资金，给人时间去做。’形式各不相同，但真正做得好的公司，比如 Vercel，他们每周五都会做一次黑客松演示。一开始只是某个小团体的娱乐活动，后来跨职能部门的其他人也说：‘嘿，我们能加入吗？’然后大家就只是去创造。每一次这些人展示出的东西都像是“创业公司杀手”，然后这些东西就被搁置了。我觉得这太不可思议了。我们只在那里待了一个星期，就是他们的一次小型公司团建。所有人都参与——比如销售团队，他们会像典型黑客松那样拉一个设计师或工程师入伙；如果没有，他们也会自己硬上。到了周五，大家抢着排队去演示。这真的是全公司范围的事。我也想做到那样。这很难，对吧？我很理解那些说‘老兄，我哪有时间做这个’的人。你会经历两个极端。我做过一些实验，让我觉得‘天哪，这效果太好了，效率太高了。’然后我也做过另一些实验，让我觉得‘我花了一整晚，想在 Quiver API 之上做一个自定义图标生成平台，但就是没能让它按我想要的方式跑起来。’我会想：‘我刚才烧掉了多少钱，还有我生命中的六小时？’根据我们的调查，人们不采用这些工具的三个原因是比较接近的。那是一道多选题：他们觉得输出还不够好；有些情况下太贵；还有就是没时间。说白了，大家都有全职工作。如果公司不给他们留出空间，很多人可能已经在周末加班加点把东西赶出来，就为了周一上班时看起来像是上周就完成了。所以这方面我非常理解。这也是为什么我把责任放在领导者身上，让他们说：‘嘿，你必须创造时间。’如果成本实在太高，我还见过另一种做法：改成每月或每季度一次黑客松，给人们大约一周时间。你可以采用类似“卡马克周”的方法：‘嘿，集中工作时间去想一个点子，然后我们来做演示，这会是一个很好的经验总结机会。’不过我要提醒你一点：如果你真的让大家在某种黑客松里高强度冲刺，把脑子榨干，之后一定要安排休息日来恢复。因为认知负担非常大。你有这种感觉吗？你会不会被搞得认知枯竭？

I think the number one is that the company is making room for the experimentation, and it's expensive. It's expensive in token cost. It's expensive because you go down rabbit holes that are throwaway work a lot. Leaders are saying, 'No, I'm going to unblock it. We're going to throw spend at it. We're going to give people time for this.' And it comes in different flavors, but the ones who are really doing it well, like at Vercel, they do a Friday hackathon presentation every Friday. And it started as a fun thing for a very specific group. Then other people from cross-functional departments were like, 'Hey, can we get in on that?' And people just create things. They're presenting startup killers at every single one of these things. And those things just get shelved. It's wild to me. And we were there for just one week, like a little company getaway that they did. And everybody—the ideas were like the sales team, they would go out, you know, like a typical hackathon, they'd recruit a designer or an engineer. And if they didn't have one, they'd just do it anyway. And people were fighting to get in line to present on that Friday. It's so company-wide. I'm trying to do that. It's hard, right? I feel for everybody that's like, 'Man, when do I get time to do that?' You get both ends. I've done experiments where I'm like, 'Dude, I can't believe how well that worked. I can't believe how efficient that was.' And then I've done other ones where I'm like, 'I just spent an entire night trying to build a custom icon generation platform on top of the Quiver API, and I just didn't quite get it to work the way I wanted to.' And I'm like, 'I wonder how much money I just lit on fire, and six hours of my life.' The three reasons people are saying from our survey that they're not adopting the tooling are pretty tied. It was a multiple-choice question, and it was that the outputs they didn't feel were good enough yet, it was too expensive in some cases, and that there was no time. People are frankly too employed. And if their company's not making room for them, people are already spending their weekend probably trying to get stuff out the door just so they can show up Monday and look like it got done last week. So on top of that, I feel for that a lot. And that's why I put the onus on leadership to be like, 'Hey, you have to make time.' And if it's really expensive, another way I've seen this done is they'll just do the hackathon once a month or a quarter and give like a week to people. You can do kind of a Carmack week method where it's like, 'Hey, go heavy on work time on an idea, then let's present it, and that could be a really cool lessons-learned opportunity.' The one thing I'll call out to you about that is, if you do make room for people to go heavily in some sort of hackathon, like just squeeze their brain in the base, you've got to attack on that day of rest afterwards, because it's cognitively a lot. Do you feel that? Do you get cognitively wiped out?

有。我也不知道。我觉得我坐上了自己那条奇怪的创造力过山车，有时候我根本不知道主导因素是什么。

Yes. I don't know. I think I'm on my own weird creativity roller coaster, and sometimes I don't know what the dominant factors are.

你这是卡马克年。你就沉浸在其中。

You're having a Carmack year. You're just in it.

是啊。我可以连续四五天非常拼命，简直像在飞。我做出来的每样东西感觉都很好，效率高得惊人。然后到了某个周二下午，我会说：‘我就出去漫无目的地走走，因为我根本不知道该拿自己的人生怎么办。’这个现象还有另一面：我刚才讲的、以及我看到的那些东西，对像我这样经常换赛道的人帮助更大。而当你面对专注于单一任务的专家时，他们也能从中获得很多，但某种程度上也许会更快碰到天花板。我不确定。也许这跟智能体式、系统性的东西更相关，但我也开始这么怀疑了。

Yeah. I can go so hard for like four or five days where I'm just flying. Everything I'm putting out feels good. I'm incredibly productive, and then it'll be a Tuesday afternoon and I'm like, 'I'm just going to go walk aimlessly because I don't know what to do with my life.' There is another piece of this that says all the things I've just talked about—and have seen—skew more helpful to people like me who are changing lanes a lot. Whereas when you get to the specialist focusing on individual tasking, you can get a lot out of it, but at some point there's maybe a faster ceiling. I'm not sure. Maybe it's more agentic, kind of system related, but I'm starting to wonder that too.

我喜欢这个，因为我每天都会切换赛道，你知道吗？

I love this stuff because I switch lanes daily, you know?

我几个月前就发过推文，但有效多任务处理的能力——在很高的水平上并行开展工作——是这种 AI 世界里运作的核心技能之一。我总是同时进行多种不同类型的任务。我同时处理不同的工作树、不同的线性问题。也许我在同一时间既做设计又写代码，然后在两者之间来回切换。我内部的工具是——所有事情我都用 Conductor，他们有一个小小的 choo-choo 声音，当智能体完成工作时可以打开。我就是从一个 choo-choo 移动到下一个 choo-choo。这就像一场打地鼠小游戏，这些 choo-choo 在做各种不同的事情。我的整个生活就是：choo-choo，option 空格键，跟电脑说话。choo-choo，option 空格键，跟电脑说话。我整天就这么干。我还想说点别的，因为我们在聊这些的时候——我没什么机会跟太多有我们这种视角的人交谈。我们有这种一万英尺的视野。我觉得人们不会意识到。我昨天试着数了数今年我跟设计师和设计负责人进行了多少次对话，已经差不多 200 次了。太疯狂了——我每天都在进行多次对话。我要告诉大家的是，当我和你聊天时，我喜欢聊这些，因为我们可以比较很多笔记，而这是我跟其他很多人没法做的。但我也要说，现在仍然非常早。它正在快速加速，但如果你听到我们这样聊，然后你心想“这真的让人应接不暇，让我有很多错失恐惧（FOMO）”，那你完全可以等着。我真的觉得你现在可以等等这些东西，尤其是所有这些智能体式的东西，因为总会有人推出一个工具，最终让我想从我的 open-claw 设置中退出，因为它做得太好了，运行得很好，而且最重要的是切换成本真的很低。那可能是进入这些东西的好时机。你不需要像我现在这样担心那些拧螺丝的麻烦。我觉得总有一天，进入它会变得认知上非常容易——这就是我的想法。

I tweeted that months ago, but the ability to effectively multitask — being able to parallelize work at a really high level — is one of the core skills of operating in this AI world. I always have multiple different types of tasks going simultaneously. I'm working on different work trees, different linear issues at the same time. Maybe I'm designing and coding at the exact same time and just bouncing in between. My tooling inside is like — I use Conductor for everything, and they have this little choo-choo sound you can turn on when an agent is done working. I just move from one choo-choo to the next choo-choo. It's like a little game of whack-a-mole, and the choo-choos are doing all different kinds of things. That's my entire life: choo-choo, option-spacebar, talking to my computer. Choo-choo, option-spacebar, talking to my computer. I just do that all day. I want to say something too, because we're geeking out on — I don't get to talk to a ton of people who have the perspective that we have. We've got this 10,000-foot view. I don't think people will realize. I was trying to count yesterday how many conversations I've had with designers and design leaders this year, and it's already in almost 200, I think. It's been wild — daily I'm having multiple conversations. The thing I'll tell people is that when I talk to you, I like geeking out about this because we can compare a lot of notes that I can't compare with a lot of folks. But I also will say this is still really early. It is accelerating pretty quickly, but if you hear us yapping about this and you're like, 'This is really overwhelming, this makes me have a whole bunch of FOMO,' you are fine to wait. I really think you can wait at the moment for some of this, especially all this agentic stuff, because someone's going to come out with a tool that is eventually going to make me want to off-board from my open-claw setup because it's just so well done, it works really well, and most importantly the change cost is really cheap. That's probably going to be the great time to get in on this stuff. You're not going to have to worry about all the tightening of the screws the way I have to do right now. I think there will come a time where it'll be really cognitively easy to get into it — that's my thinking.

过去一年，Dive Club 让我非常清楚地看到一件事：设计的实践正在改变，而旧有的反馈流程在当今世界已经不太管用了。这就是为什么我很高兴地宣布 InFlight 正式进入公开测试阶段。这是我一直想要的反馈工具，它为 AI 速度的世界而生。我可以分享我的原型，在视频演示中提供背景，InFlight 让我很容易获得继续前进所需的精确反馈——无论是投票选择方向，还是可能给新创意开绿灯。所有这些都可以通过一个链接完成，我可以放到 Slack 里，或者分享给核心用户来测试新原型。我每天都在用 InFlight，它彻底改变了我的分享方式。所以我很期待大家试用这个产品，如果你想聊聊它，给我发邮件：rid@inflight.co。

One thing that Dive Club has made abundantly clear to me over the last year is that the practice of design is changing, and the old process of getting feedback just doesn't quite cut it in today's world. That's why I'm excited to announce that InFlight is officially in open beta. It's the feedback tool that I've always wanted, and it's built for a world that moves at the speed of AI. So I can share my prototypes, give context in video walkthroughs, and InFlight makes it easy to get the exact feedback that I need to move forward — whether it's voting on directions or maybe even getting the green light to ship a new idea. And all of this is available in a single link that I can drop into Slack or maybe even share with power users to test out a new prototype. I use InFlight every day, and it's totally transformed the way that I share work. So I'm excited for you to try the product, and if you ever want to jam about it, just email me at rid@inflight.co.

那我们就从那里拉远一点，利用这个制高点和视角，毕竟我们确实与业内人士、领导者、团队等进行了大量对话。所以，也许回到你做过的一些现场团队访问。我很好奇，尤其是考虑到你谈到 open-claw 的那些东西。我知道现在很多听众都感到一丝压力，也许还有担心自己落后了。你怎么适应？当感觉你可以同时朝每个方向前进时，你会把自己指向哪里？那么，你与那些团队共度的时间是否塑造了你对设计师未来角色的看法，但更具体地说，我们应该投资在哪里，作为专业人士，在接下来的几个月和几年里，我们将带来什么？

Let's zoom out from there then and tap into that kind of vantage point and perspective, given the fact that we do have a lot of conversations with people in the industry and leaders, teams, that kind of thing. So maybe going back to some of the on-site team visits that you've done. I'm curious, especially given what you're talking about with the open-claw stuff. I know that a lot of people listening right now feel this little tinge of pressure and maybe this worry that they're falling behind. How do you adapt? Which direction do you point yourself when it feels like you could go in every single direction simultaneously? So, did your time with those teams do anything to shape the way that you think about the future role of the designer, but even more specifically, where we should invest and what we will even bring to the table as professionals in the months and years moving forward?

关于这个我有很多话想说。所以，在这些看法里，有来自一线设计师的心声。有很多好的看法，也有一些非常——有减速主义者，有害怕的人，还有加速主义者。另一方面，就是人们实际在专注什么？他们在变成什么样？我看到的是，越来越多的人发现自己被赋予责任，并不得不去领导某些东西——无论是智能体式装置的落地，还是人员，或者两者都有。我希望看到这些角色开始有与之匹配的报酬。那是我 2024 年的假设。有这样的画面：随着工具变得更好，工作容量开始打开，比如你一天能做更多事。我认为其中有一个未被触及的认知维度，因为那是真的。但你可以做得更多。就像你有更多时间，更多能力。你在那个开放空间里堆叠更多东西，报酬应该增加。这就是假设。我实际上想在本季度做一项调查，追踪进展情况。对于团队——我认为人们应该在哪里投资技能，或者他们在关注什么——就是能够建立意图，并能放大缩小。这种在不同高度之间移动的能力，不仅是产品思维和交互时刻，还有工作流。Ben Blumenrose 对什么让一个人成为 AI 原生，给出了一个很好的答案。它不是关于工具。它不是关于——当然里面有某种能力，但对他来说，关键是你如何对待“如果？”这个问题。我觉得这个框架才是我更希望设计师关注的。因为有人会问我，“嘿，我现在该怎么做？我该创建一个项目吗？我该玩玩 Claude Code 吗？”我会说，先屏蔽噪音，想想你手头的问题。每个设计师都有的一个问题就是空白画布问题。我觉得这是进入这些东西的 Hello World 项目。然后去弄清楚如何为你自己解决空白画布问题。也就是说，我们会做疯狂八（crazy eights）。

I have so much to say about it. So in there, there's the sentiment you hear from working designers. There's a lot of sentiment that's good, and there's some very — there's the decelerationists, there's people who are afraid, and there's the accelerationists. Then on the other side, it's like what are people actually focusing on? What are they becoming? What seems to be happening is more people are finding themselves being given responsibility — that's what I'm seeing — and having to lead something, whether it's the agentic installation or people, or even both. I want to see those roles starting to have compensation to match that. That was kind of my hypothesis from 2024. There's this visual that as the tools get better, the capacity for work starts to open up, like you can do more in a day. And I think there's a cognitive component of that that wasn't addressed, because that's true. But you can do more. It's like you have more time, you have more ability. You stack more things inside of that open space, compensation should increase. That's the hypothesis. And I actually want to run a survey this quarter tracking where that's at. For teams — where I think people should invest skills or where they're looking at it — it's just like being able to build intent and zoom in and out. The ability to move between altitudes, not just product thinking and the interaction moment, but also workflows. Ben Blumenrose had a really good answer to what makes someone AI native. And it's not about the tools. It's not about — there is some competency in there, sure, but to him it was: what do you do with the question 'what if?' And I feel like that framing is really where I would rather have designers focus. Because I'll have some people who are like, 'Hey, what do you recommend I do right now? Should I create a project? Should I play with Claude Code?' And I would just say, tune out the noise for a second and think of something that you have as a problem. And a problem every designer has is the blank canvas problem. It's the Hello World project, I think, to get you into this stuff. Then just go and figure out how to solve the blank canvas problem for yourself. Meaning we would do crazy eights.

你知道，我们会把一张纸折起来，做八分钟的快速头脑风暴，每个点子一分钟，来解决某个界面问题。在现有工具下，这个做法对应什么？去试几个工具，搞清楚空白画布问题是什么。也许你打开 Paper，用新工具做你熟悉的事情。或者你打开 Claude Code，用 vibe code 写出八个快速变体。或者你更进一步：“嘿，我怎样把 Claude Code 连到 Papers MCP？然后我怎样在 Claude Code 里制定计划，再在 Papers MCP 里执行？这样我打开 Paper 就能看到八个可能很烂的版本，然后直接判断哪些是垃圾、哪些有意思。”就用你感兴趣的方式去做。不要去看 Tommy 的 newsletter，也不要去看 Dive Club 的 newsletter，然后照着做。把那些都屏蔽掉，挑几件你觉得好奇的东西。练习思考，然后去试。开始建立你自己的小洞察库，记录你在学什么、你工作流里喜欢什么。

You know, we would fold a piece of paper and do quick crazy eights in eight minutes, one minute per idea, to solve some interface problem. What is the version of that with the tooling available? Go and explore a couple of tools to figure out what the blank canvas problem is. Maybe you open up Paper and use the new tool to do stuff you're already familiar with. Or maybe you open up Claude Code and vibe code eight quick variations. Or maybe you go even deeper and you're like, 'Hey, how do I connect Claude Code to Papers MCP? And then how do I actually create a plan in Claude Code and have it executed in Papers MCP? So that I just open up Paper and I have eight probably shitty versions of this instance that I can just decide what's trash and what's interesting.' Just do that in a way that's interesting to you. Don't look for Tommy's newsletter or the Dive Club newsletter to figure out what you should do. Tune it all out and pick up a couple of things that you found curious. Practice thinking and just try it. Start to build your own little database of insight around what you're learning and what you like about your workflow.

因为每个人的工作流——这是我见过很多次的一点——都不一样。如果我们把每个人的工作流截图放出来，

Because everybody's workflow — and this is the one thing I have seen a lot — is different. If we were to put screenshots up of everybody's workflows,

对。

Yeah.

这些新冒出来的工作流，形状大小各异，奇形怪状，而且现在都非常个性化。即使在同一个组织内部也是如此。

that are emerging, they're all shapes and sizes, oblong and weird, and they're just very custom right now. Even within an org.

对，即使在一个组织内部。

Yeah, even within an org.

组织内部工作流之间的差异——比如我一直在打大量客户电话，兄弟，这真让人震惊。显然每个人都有自己接入创造力的流程和方式，但人们在使用完全不同的工具、工作方式和思维方式。创意设计流程的分化已经变得极其极端。这很混乱，但我觉得那些分享自己工作流的人、团队和个人，里面有大量价值。而且我认为要记住，现在这确实是个好活动。你不必给它贴上“嘿，这是我的工作流，你应该这么做”的标签。对吧？我现在想把“应该”这个词从词汇表里划掉，只去比较工作流、比较形状。因为我们最终会在这些事情上收敛出一些更好的实践，但我们还不知道那是什么。所以现在，如果你想参与这类工作，最好的事情就是试着找到适合你的工作流，然后把大概 80% 的时间花在那些能完成工作的熟悉事情上，再用一部分时间给自己留出探索新工作流的空间。所以对于那些我说“你可以等”的人，重点是：我们终会收敛。我们会在某些东西上收敛。如果你有这个奢侈，还不需要参与进来，你可以等，那时候很可能会有“这些是更好的工作流”的集结。

The delta between workflows within an org — like I've been doing a lot of customer calls, and it's shocking, man. Obviously everybody had their own process and means to tap into creativity, but people are using fundamentally different tools, ways of working, ways of thinking. The spreading apart of the creative design process has gotten so extreme. It is chaotic, and I think there's a lot of value in the people, the teams, the individuals who are sharing their workflow. And I think it's important to remember that that's a good activity right now. You don't have to attach language to it like, 'Hey, here's my workflow, you should do this.' Right? I like to strike 'should' from the vocabulary right now and just compare workflows, compare shapes. Because we will eventually converge around some better practices around this stuff, but we don't know what that is yet. So right now, the best thing you can do — if you want to get involved in this type of work — is to try to find workflows that are good for you and spend, like, 80 percent of your time on the familiar stuff that gets the work done, and then some percentage of the time where you make room for yourself to explore a workflow. And so the point for the people who I said 'you can wait' is: we're going to converge. We are going to converge around stuff. If you have the luxury and don't have to get involved yet, you can wait, and there probably will be a concentration of 'these are the better workflows.'

我觉得我们的旅程有些相似之处，但可能对听这个节目的普通设计师来说也很独特，那就是我们都在分发上做了投入，把想法发出去，看看它会带我们去哪儿。至少对我自己来说，我有点像搭上了 Figma 加入大量技术特性的这波浪潮。说实话，当时有一种新的设计类型，感觉更像工程。我非常容易就能上手捣鼓。我很好奇，就直接去做了。我尝试了每一种不同的小策略，不断把想法发出去。不是“你该这么做”，而是“嘿，我试了这个，真的很酷，我明天还要再试一次”。我觉得我们现在正处于这样一个时刻：对于那些想在设计社区里成长的人来说，这个窗口又彻底打开了。人们非常渴望看到别人的工作流，看到别人正在做的百分之百不同的实验，以及别人获得的不同经验。如果你只是出现、尝试、保持好奇，然后把 Twitter 当成一本记录你正在学和正在做的实验的小日记，那么现在就是收获公开好奇红利的时候。

I think something that makes our journeys similar, but also unique maybe to the average designer who's listening, is we both made this investment in distribution, putting our ideas out there, kind of seeing it where it would take us, and at least for myself, I kind of rode this wave of Figma throwing in a bunch more technical features. There was this new type of design that felt a little bit more like engineering, honestly. And it was really easy for me to just tinker. I was curious. I just went for it. I just tried every little different tactic and kept putting ideas out there, and it wasn't like 'do this.' It was more, 'Hey, I tried this. This was really cool. I'm going to try this again tomorrow.' And I think we're in this moment in time where that opportunity is wide open again for people who do kind of want to grow within the design community. There's so much hunger and desire for people to just see workflows and 100% different experiments that people are running and the different learnings that others have. And if you just show up, try things, be curious, and then treat probably Twitter like a little journal of the things that you're learning and the experiments that you're running, like this is the moment in time right now to reap the rewards of being publicly curious.

我对最近几个比较新的设计师说了非常具体的建议，他们问我：“嘿，我该怎么做？”我说：“好，就做我刚才说的那件事——挑一个项目，做一个实验，去 Twitter 分享。你不必装得像大师一样。这不是关于大师。这其实是关于……”

I've been very prescriptive to the last couple of people who are newer designers and they asked me like, 'Hey, what should I do?' I said, 'Okay, do the thing I just said there, like pick a project, do an experiment, go to Twitter, share it. You don't have to act like you're a guru. Like this is not about guru. This is about'

人们会嗅出来。

People will sniff.

其实，其实别这样。对。而是要公开构建，然后说：“我做了什么。”因为关键在于：真正能火的东西——如果你想谈纯算法最大化——就是去做一些非常视觉化的 vibe code，某种奇怪而独特的东西，比如 3D 图，或者某个 3D 时空的东西，然后分享出去。现在用 vibe code 做内容太容易了。但如果你想让它变得有用——这也是我推荐的——就去创建一个针对你想解决的项目工作流的具体东西，然后从中捕捉视觉产物。里面有某个瞬间特别有视觉趣味，你就分享出来，再附上几点你学到的经验，你就能收获关注。现在这对客户项目、对销售线索开发都非常有效。

In fact, in fact, don't. Yeah. And be building public and just say, 'Here's what I did.' Because here's the thing: the things that really pop off — if you want to talk about just algo maxing — would be to create vibe code something really visual, something weird and unique, like a 3D graph or some 3D space-and-time thing, and just share it. It's so easy to vibe code for content now. But if you want to make it useful, and that's what I would recommend, create something that is specific to a workflow you want to solve for a project, and then just capture the visual artifact out of that. There's some moment in there that's really visually interesting, and just share it with a few bullet points on what you learned, and you would just farm. That would work so well right now for client work, for lead gen.

我不知道你是否也注意到了——我非常想听听你的看法——但我开始注意到，特别是在就业市场上，有一种需求，倒不一定是写在纸上的需求，而是对那种能帮助整个组织提升工作方式和 AI 系统的设计师的需求。那种能开路的人。如果你能把自己定位成那样，无论是在面试中，还是仅仅在你的 Twitter 简介里，这在今天的就业市场都是一个巨大的加分项。你见过这种情况吗？

I don't know if you're seeing this too — very curious to hear your perspective — but something I'm starting to notice, specifically in the job market, is that there's demand, not on paper necessarily, but demand for designers who can help level up the rest of the org and the way they work and their AI systems. The person who can just blaze the trail. If you can position yourself as that, whether in an interview or even just on your Twitter profile, it's a massive green flag in today's job market. You seen that at all?

对，真正有意思的事情仍然是内部工具建设。如果你不是组织里指定的领导者，要想让自己被那样倚重，最好的方法就是做一点 Brandon Engineer 在 Vercel 做的事。找一个跨职能的用例：市场部需要设计，但你没有时间、也不想花时间去做的任务，然后做一个内部工具。我记得他们说 Ramp 内部已经用 vibe code 做了 1500 个工具，对吧？那问题是，其中有多少是一次性的？嗯，大部分是。

Yeah, the thing that's been really interesting again is internal tool building. If you're not a designated leader at the organization, the absolute best way to find yourself being leaned on in that way is to do something like Brandon Engineer did at Vercel. Find a cross-functional use case where design is needed by marketing for a task you don't have time for, you don't want to spend time on, and build an internal tool. I think they said they had 1,500 vibe-coded tools inside Ramp, right? And the question is, well, how much of that is throwaway? Well, most of it.

大部分东西都是用完即弃的。问题是：内部采用到底怎么发生？其实和其他产品一样：只要它有用，有人试了，就会开始传播——「哦，某某为这个做了个东西」。你装了一个新东西，就成了那个被安装对象的主题专家。大家会开始看着你说：「嘿，那真有意思，我该怎么做？」然后他们会为你腾出时间。如果你在一家没太多时间搞这些的公司，这几乎就是一条可行路径——只要你有能力以那种方式使用工具，这也是另一个问题。那是一种很棒的做法，也是我会推荐的做法。

Most of it is throwaway. The question was, how does internal adoption actually happen? Well, it happens just like anything else with a product. If it's useful, someone tries it, and then it starts to spread: “Oh, so-and-so built a thing for this.” You're installing something new, and now you're the subject matter expert on what's being installed. People start looking at you and saying, “Hey, that was really interesting. How do I do that?” And they'll start making time for you. That's almost a path if you work at a company that doesn't have a lot of time for this stuff, but you do have the ability to use the tooling that way—which is another issue too. That's a fantastic way to do that, and that's what I would recommend.

我想到了好几期节目里的例子：在 Notion、Atlassian、Stripe 搭建早期原型实验场的人，在组织内部一下子声名鹊起。突然之间，不同设计师都开始依赖他们。这里有太多东西要搞明白，而设计师就是解决问题的人。想到我们日常角色中要解决的很多问题可能更多是内部问题、而不是面向用户的问题，很有意思。而且现在做这些事的空间也更大了，因为我们可以更快完成面向用户的任务。

Multiple episodes come to mind where people built the early prototyping playgrounds at Notion, Atlassian, and Stripe—they were catapulted internally in the org. It became a big deal all of a sudden; different designers were relying on them. There's just so much to figure out, and designers are problem solvers. It's interesting to think about how a lot more of the problems we'll solve in our day-to-day roles might be internal rather than user-facing. There's actually more bandwidth for that now too, since we can accomplish the user-facing tasks much more quickly.

人们会问：为什么要造出取代你自己的工具？你只是在替换自己，这是在坑自己。但在 Ramp，有两件事让我印象深刻。他们真的很希望设计师专注于在关键摩擦点上创造 10 分满分的体验。你不需要去给市场部做博客配图——作为产品设计师，那可能不是你时间的最佳用途。

People ask, why would you create tools that replace yourself? You're just replacing yourself; it's a self-own. But at Ramp, two things stood out to me. They really want designers to focus on creating 10 out of 10 experiences at the friction points that matter. You don't need to be creating the blog post image for marketing—that's probably not the best use of your time as a product designer.

另一件让我特别感兴趣的事是：那些真正在内部被采用的工具看起来真的很酷。这有点奇怪，因为我每次做内部工具——我记得在 Quantcast 有一个，当时客户经理需要点东西——目标就是让它能用。就像 Bootstrap，一堆难看的东西绑在一起，能跑就行。即使后来它被升级了，也还是低于标准。但现在我看 Ramp 里的一个工具——我记不清具体是什么了，反正不是 Glass。Diego 和 Simon Corey 一直在做那个首字下沉字母，你每天都会收到一份摘要。那个首字下沉字母就像一幅挂毯，带着老式文艺复兴字体，真的非常漂亮。我当时想，一个内部工具居然投入了这么多细节，太疯狂了。然后是 Shopify 的 artifact 工具：那条推文称它是年度最佳设计产品之一，可它只是一个内部工具。之后的几周里我收到了大量私信，人们问：「你能直接为 Invite 做这个吗？这就是我们想要的 Invite。」我心里想的是：能。但另一部分我又觉得：我不知道自己能不能设计出那么好的 UI——说真的，那个内部工具可能超出了我的能力范围。这太不可思议了，因为你说得对。如果一个人简历上写着「过去几年在 X 公司做内部工具」，你的大脑会立刻觉得：他们可能没什么视觉能力，大概只会做灰盒功能界面。但现在情况不再是这样了。

The other thing that was really interesting to me was that the tools that do get adopted internally look really cool. That's weird, because every time I've worked on an internal tool—I remember one for Quantcast where our account managers needed something—it was just about making the thing work. Bootstrap, a bunch of ugly things strapped together, and it just worked. Even when it got leveled up later, it was still subpar. But now I'm looking at a tool at Ramp—I can't remember what it was; it wasn't Glass. Diego and Simon Corey had been working on this drop cap letter, and you'd get a summary every day. The drop cap was like a tapestry with old Renaissance lettering, and it was really beautiful. I thought it was crazy that that much attention to detail went into an internal tool. Then there's Shopify's artifact tool: there was a tweet calling it one of the best design products of the year, and it was just an internal tool. I received so many DMs in the weeks afterwards—people asking, “Can you just build this for Invite? This is what we want for Invite.” In my head I was like, yes. But the other part of me was like, I don't know if I can design a UI that good—legitimately, that internal tool might just be beyond what I'm capable of making. Which is insane, because you're right. If you saw someone whose resume said they'd been building internal tools at X company for the past few years, your brain immediately assumes they probably don't have visual skills. They're probably building gray-box functional UI. But that is not the case anymore.

我保持谨慎乐观，并不是彻底的全情投入。有些人认为应该把一切都移开，让加速尽可能快地发生，因为我们必须尽快穿越那段痛苦的中间地带。我听过这种观点，而且大体上认同。但我仍然觉得不要开到时速 100 英里，可能 60 英里就好。按限速行驶，并在过程中留意我们看到的景象。不过我们还是要想办法到达那里。我认为，到达那里是值得为之努力的目标。

I'm cautiously optimistic. I'm not full-blown all in. There are some people who think we should move everything out of the way and let acceleration happen as quickly as possible, because we need to get through the painful middle as fast as we can. I've heard that take, and I mostly align with it. But I still think don't go 100 miles an hour—go maybe 60. Go the speed limit, and pay attention to what we're seeing as we get there. But let's try to get there. I think getting there is a worthwhile cause.

好，我们花点时间聊一位非常具体的听众。他在中西部一家中型代理公司工作，已经待了三四年，之前还零零散散做过一些自由职业。他的简历不会让谁眼前一亮，但他相信自己有能力迈出一大步，拿下一个人人向往、也算能让他崭露头角的职位。帮他想想该把方向指到哪里。除了日常工作，他时间有限，到底应该把时间花在哪些事情上？有哪些作品、流程或知识，能让他武装自己，迈出这一步，进入大联盟？

All right, let's talk to a very specific listener for a second. They're working at a Midwest mid-tier agency. They've been there for three or four years, and did some random freelancing before that. They're not going to win people over with their resume, but they believe they have the skills to take a big leap and land one of those roles that people care about—the kind that puts them on the map, so to speak. Help them think about where to point. What are the ways they should be spending their time? You only have so much time outside of day-to-day work responsibilities. What artifacts, processes, or pieces of knowledge can they use to equip themselves to take that leap and play in the big leagues?

我认为现在你能创造价值的最重要方式——而且我真的认为，不是每个设计师都想成为系统设计师，不是每个人都想思考系统，这点我理解。确实仍有很多视觉专精的人会被重视。但我觉得现在最关键的，是去帮助跨职能的同事更容易地完成设计工作。让你的市场团队不需要你也能做出博客配图。去构建一个赋能跨职能伙伴的设计职能。我真的不认为你会亲手把自己淘汰——不会发生这种事。你建立这些流程之后，他们不会说「好的，谢谢，再见」。恰恰相反，他们会发现越来越多事情需要这类系统思维。会有更多事情需要真正的产品级贡献，而那些不是跨职能伙伴该聚焦的。于是你会有更多时间去做那些事情。当你有了更多时间，我就会想到像 Andrew Wendling 这样的人，他是 Iverson Studio 的一位出色设计师。他现在活跃在疯狂的生成式图像/视频世界里，并把那些探索和编程项目结合起来，做出非常丰富的体验。

I think the biggest thing you can do to add value right now—and I really think not every designer wants to be a systems designer, someone who thinks about systems; I can appreciate that. There's definitely still a world where a lot of very visual-specific folks are valued. But I think right now, the biggest thing you can do is go in and try to help cross-functional people do more design work easily. Enable your marketing team to create that blog image without needing you. Create a design function that enables cross-functional partners. I truly don't think you're going to cannibalize yourself—that's not what's happening. You're not going to build these processes and then they'll say, “All right, thanks, see you.” Instead, they're going to find more and more things that need that sort of systems thinking. There are going to be more things that need actual product contributions—things that cross-functional partners shouldn't be focusing on—and you're going to have a lot more time to do that. When you have that time, I think of someone like Andrew Wendling, a wonderful designer at Iverson Studio. He's living in this wild generative image/video world and combining that with coding projects, getting really rich experiences.

Amelia Wattenberger 分享过一些博客文章，它们是不同生成式图像之间交互组合的产物，还有一些关于界面工作未来的、思辨性很强的深度文章。你会发现，这类东西之所以能出现，是因为你现在有时间了——因为你启用了这些系统，它们把你不得不做的工作中的辛苦劳作去掉了。

Amelia Wattenberger shared some blog posts that are these combinations of interactions made between different generative images, along with really thoughtful, philosophizing posts about the future of interface work. And things like that become possible because you now have time — because you've enabled these systems that remove the toil from what you have to work on.

我想就这点再追问一下。因为你可以在你想离开的那家老公司里，做出你能想象的任何最有影响力的系统。但要在这套技能之上想办法在市场中脱颖而出、证明你的价值、并用一种让人愿意在你这个人身上停留超过 10 秒的方式去做，那就是另一整套挑战了。所以，你会怎么从那里往前走？因为我觉得，事实上那并不像你说的那种搞怪机器在 Twitter 上被人转起来那么容易卖。

Let me push on that a little bit, because you could make the most high-impact systems imaginable at that old company you're trying to leave. It's a whole other set of challenges to figure out how to stand out in the market on top of that skillset, demonstrate that value, and do it in a way that gets people to spend more than 10 seconds on you as a person. So how would you go from there? Because I think the fact is that's not as easy to sell as the kind of goofy machine you were talking about that gets played on Twitter.

我没有完美的答案。我不想装成我今天有多懂怎么拿到 offer。但我告诉你我的意思。有些实习项目仍然有白板题环节，但在此之前，他们会做 20 分钟的现场写代码练习。我觉得那很有意思。他们就是想去测试这些能力。对，他们说他们不想看 Figma，他们想说‘来，现场把它码出来’，而且那是一个产品设计岗。现在有人在想办法弄清楚如何评估 AI 熟练度。我觉得 AI 熟练度最终会变成一种像空气一样自然存在的东西。我不觉得我们会永远把这个词挂在嘴边。但我觉得现在它还是重要的关键词，因为很多公司都在说：对，我们想要一个来了就能思考的设计师。我们希望有人知道该怎么面对‘如果……会怎样’这个问题，而且不会给出那种像是 2012 年、2015 年、2018 年的感觉的答案。这个问题似乎正在由你展示的东西、你分享的东西来回答。社交媒体就是当下的新发现引擎。不管好坏，你就是要用这种方式来证明你是一个在实验的人，在以一种很多其他人没有的方式探索‘如果……会怎样’这个问题。

I don't have the perfect answer. I don't want to act like I'm a subject-matter expert on how to get hired today. But let me tell you what I'm saying. Intern programs that still have whiteboarding — but before that, they do 20-minute exercises of live coding. I thought that was really interesting. They want to test those capabilities. Yeah, they said they didn't want to see Figma; they wanted to see, hey, live-code this. And it was for a product design role. There are people trying to figure out how to evaluate AI fluency. And I think AI fluency is eventually going to become an ambient thing. I don't think we're going to highlight that language forever. But I think right now it is important language, because a lot of companies are like, yeah, we want this designer who can come in and think. We want someone who knows what to do with the question 'what if' in a way that doesn't feel like it's from 2012, 2015, or 2018. That question seems to be answered by what you're showing, what you're sharing. Social media is the new discovery engine right now. For better or worse, that's the way to prove you're someone who's experimenting and exploring what to do with the question 'what if' in ways that a lot of other people aren't.

假设你现在是招聘经理，正在跟我前几个问题里反复提到的那个假想人选交谈。你想琢磨出一些具体的问题或度量标准，来衡量这种 AI 熟练度。你会用什么？你希望看到哪些信号？

Let's say you're the hiring manager and you're talking to this hypothetical person I've been bringing up over the last couple of questions, and you're trying to figure out specific questions or measuring sticks to get at this level of AI fluency. What are you reaching for? What signals would you hope to see?

现在——我是说，这是我最想知道的大问题：“给我看看你的工作流程。给我看看你是怎么设计东西的。带我过一遍你做过的一个想法。”我想知道的就是这个。我想看着它，拆开它，问为什么，看看里面用了哪些工具。这其实很有挑战性，因为我觉得大家不会——‘你什么意思，我的工作流程？嗯，我去了那边，问了几个客户，然后去了 Figma，把东西做出来了。’我想知道的是：好，行，我完全理解。那有没有什么是你玩过的？它为什么成了，或者为什么没成？‘哦，是这样的，我一般会打开 Superwhisper 或者 Whisper Flow，然后我就对着它吧啦吧啦说一通，它就会变成这个样子——但我不想给你们看这个，因为那太浪费时间了。’我会说：不不不，给我看看。真的给我看看：你为什么决定用语音？‘嗯，因为我说话比较散，光靠嘴说我表达得不太好，但 AI 真的帮我浓缩了思路——然后我可以拿那些内容，填到这边的这个提示词里，我发现它真能更好地生成这个方案。’我会说：好，有意思，这下我们说到点子上了——来，深挖一下。不是每个人都有那种工作流程，但这就是我在找的东西：你有没有玩过这样的工作流程？你是怎么想的？你为什么决定那么做？如果我要找 AI 熟练度，我找的就是这个。我在找那种在游乐场里的人——不只是说你有一个用 vibe coding 弄出来的游乐场，而是你也把自己的工作流程当成了一点游乐场。你 Mac 上按 alt-tab 切出来的东西，每个星期都长得不一样。

Right now — I mean, this is the biggest question I want to know: “Show me your workflow. Show me how you design something. Take me through an idea that you've created.” That's what I want to know. I want to watch, pick it apart, ask why, and see which tools are involved. That's a challenging thing, because I don't think people are like — what do you mean, my workflow? Well, I went over here, I asked some customers, I went over to Figma, I made this thing. The things I want to know: okay, cool, totally get that. Is there anything you've played with, and why did it work or why didn't it work? Oh, well, normally I would pull up Superwhisper or Whisper Flow, and I would just be like, blah blah blah, and then it would look like this — but I don't want to show you guys that, because that's a massive waste of time. And I'm like, no, no, show me that. Actually show me: why did you decide to do voice? Well, because I'm kind of rambling, I'm not a great communicator if I just talk, but AI really helped me condense my thinking — and then I can take that and plug it into this prompt over here, and I found that it could actually generate this plan better. I'm like, okay, cool, now we're onto something — let's dig into that. And not everybody has that workflow, but that's what I'm looking for: have you played with workflows like this? How are you thinking about it? Why did you decide to do that? If I'm looking for AI fluency, that's what I'm looking for. I'm looking for people who are in the playground — not just you have a playground where you vibe coded, but you're also treating your workflow a little bit like a playground. Whatever your alt-tab is on your Mac looks different any given week.

我想把你之前说过的一句话拉回来，我觉得它跟这个有点张力，那句话就是：‘等着就好。’因为你刚才说的完全也是我说的。在我们的人才网络里，有一个在后台支撑一切的产品，叫 decimals。我有一个可以悬停在上面的东西，它会显示招聘经理在某个职位上会看的红旗和绿旗。看看这些红旗绿旗里的趋势真的很有意思。有一个红旗——到目前为止我看到的第一大红旗——就是候选人没有表现出对新的 AI 流程或工具的任何兴趣或好奇心。而你刚才描述的正好相反，对吧？那才是绿旗。如果你出现的时候能说：‘哥们儿，我一直在各个方向探索，我做了这么一件事，它在这边真的奏效了，而且我还在玩这个新工具。’招聘经理爱死这个了。他们真的爱。

I'm going to pull back the thing you said earlier, which I feel is a little bit in tension with this, which is: it's okay to just wait. Because what you just said is totally what I'm saying as well. And in the talent network, we have this product that kind of powers everything behind the scenes, called decimals. And I have this thing I can hover over that shows the green and red flags a hiring manager is looking for in any given role. And it's really interesting to look at the trends across those green and red flags. A red flag — by far the number one red flag I see — is that a candidate has not demonstrated any interest or curiosity in new AI processes or tools. What you're describing is the opposite, right? That's the green flag. If you can show up and be like, man, I've just been exploring in every different direction, and I did this one thing and it really worked over here, and I'm playing with this new tool — hiring managers love that. They love that.

关于这个平衡，我也没有完美的答案。我在努力对自己的措辞负责，不想把大家吓得陷入恐惧。还是那句话，60 英里/小时和 100 英里/小时的差别。但我确实觉得，到某个时候——因为你必须明白，这说明了你是怎样的设计师。如果你不在做实验，那说明你可能就是太忙了。你可能没有那个余裕，这是非常真实的事。你在某家公司工作——或者在 Oracle 这样的公司，那里的指导方针是：‘我们认为设计工作的未来不是那样的。’他们也许是对的。但据我所知，Oracle 非常非常反对那个方向。也许你就处在那种处境里，而那件事你也改变不了多少。而且很可能一切会没事的，因为你会搞明白的。我确实认为会有第二波采用浪潮到来。但我也确实认为，如果你现在正在找工作，而且你想脱颖而出——如果我们聊的是脱颖而出，是缩短拿到下一份工作的周期——那么去探索这些工具就是一个非常积极的信号，因为它 100% 说明你上心。不探索不代表你不上心，但如果你在探索，那 100% 说明你上心。

I don't have the perfect answer for the balance. I'm trying to be responsible with language and not just crush people into fear about this. Again, 60 miles an hour versus 100 miles an hour. I do think at some point — because you have to understand what it says about you as a designer, though. If you're not experimenting, it says that you might just be too damn busy. You may not have the luxury, and that's very real. You work at a company — or at Oracle, where the mandate is like, we do not think the future of design work looks like that. And they might be right. But from my understanding, Oracle is very, very anti that direction. And maybe that's the situation you're in, and there's not much you can do about that. And it's probably going to be okay, because you'll figure that out. I do think there's going to be a second adoption wave that occurs. But I do think if you're somebody who's looking for a job right now and you want to stand out — if that's what we're talking about, standing out, having a shorter loop to getting that next piece of work — it's a really positive signal that you're exploring these tools, because it 100% says you give a damn. It doesn't mean you don't give a damn if you're not doing it. But it 100% means you give a damn if you are.

而且有时候，人与人之间最大的区别，就是他们油箱里还剩多少油。你到底有多在乎？如果你要跟上这些东西，如果你要做出真正好的工作，你就得非常在乎。你必须非常在乎，你知道吗？我要把你放在聚光灯下。我们猜对的概率几乎为零。但假设六个月后我们再聊一次，我们再做一期播客。你觉得到那时会是更有时代热潮、而今天还没有那么深入人心的话题是什么？

And sometimes the biggest differentiator between two people is how much gas they have in the tank. How much give-a-damn do you have? If you're going to keep up with this stuff, and if you're going to do really good work, you have to care a lot. You have to care a lot about this, you know? I'm going to put you on the spot here. The chances of us getting this right are almost nothing. But let's say that we have this conversation six months from now. We run it back on an episode. What topic do you think would be much more in the zeitgeist—that we would be talking about—that isn't as top of mind today?

嗯，我的意思是，我觉得是 headless design（无头设计）的上下文层。我觉得这就是方向。我喜欢做预测，也喜欢核对我的预测，因为这能帮我打磨自己的世界观。比如，我是不是在自欺欺人？还是我确实在关注对的东西？我最近运气不错，虽然以前也不总是这样，但我觉得自己确实跟上了。我确实认为我们会看到 headless design 这个话题火起来。我现在说还太早。我们聊 OpenClaw 安装、聊 Ramp building glass，都还太早。但我认为接下来几个季度，这个趋势会真正加速。很多公司会意识到：“哦，智能体在用工具。那是什么样？又意味着什么？”Dan Shipper 刚发布了那个写作工具。

Well, I mean, I think the context layer of headless design. I think that's where this is going. I like making predictions, and I like fact-checking my predictions, because it helps me hone my worldview. Like, am I full of it, or am I paying attention to the right things? I've been on a little bit of a run. It hasn't always been the case, but I feel like I'm plugged in. And I do think we're going to see this headless design topic. I'm so early. We're so early talking about OpenClaw installations and like Ramp building glass. And I think over the next few quarters, we're going to see that really accelerate. I think a lot of companies are going to realize, 'Oh, agents are using tools. What does that look like and what does that mean?' Dan Shipper just launched that writing tool.

没错。

That's right.

而那是一个非常有说服力的案例：和一个一直保持专注的觉醒 AI 并肩工作。它会说：“嘿，其实你应该试试这边。”或者说：“其实有个新洞察，也许你可以考虑一下。”我觉得第一阶段的讨论就是那个公文包的安装，也就是在工具之间移动。在那之后，我觉得就会变成：“好，人们实际上是怎么和一个永远在线、就在你工具里的 AI 一起工作的？”这会是真正的讨论话题。对有些人来说，这听起来像科幻。其实我们已经到了。我们确实已经到了。只是创建和管理那个公文包还有太多摩擦。

And that was a really compelling case for working alongside an awakened AI that's just paying attention. Like, 'Hey, actually you should do this over here,' or, 'Actually this insight came in—maybe you can consider this.' I think the first conversation is the installation of that briefcase, moving tool to tool. And after that, I think it's going to be like, 'Okay, how are people actually working with an AI that is always on, in the tool with you?' That's going to be a real conversation that comes up. And that sounds like sci-fi to some people. Like, we're there already. We are there. There's just a lot of friction to creating and managing the briefcase.

没错。现在的上手体验糟透了。

Correct. The onboarding is terrible right now.

现在的上手体验还不行，你知道的，但这会很快被解决。你想一想公文包内部的设计：文件夹在哪里？什么放哪里？那些小文件纸张的分类体系是什么？我从来没用过公文包，但我想象里面会有一些小隔层。它的设计责任会落在设计师身上，和其他人一样多。

The onboarding isn't there yet, you know, but that's going to be solved very quickly. You think about the design of the interior of the briefcase: where are the folders? What goes in where? What's the taxonomy of the little file paper? I've never had a briefcase, but I would imagine there are little compartments inside of it. The design of that is going to fall as much on designers as anybody else.

百分之百。这是一类完全不同的待解决问题。

100% It's a whole different type of problem to solve.

是啊，这也是为什么你会看到 Google 还在发布 design.md 文件。你会看到很多计划试图把它标准化。比如，你如何把意图标准化？我们正在朝那个方向走，一年之后看起来可能会非常不同。另外，你如何在整个组织里分享沉淀下来的经验教训和最佳实践？因为我们一开始就说过，我们根本都不读自己的 Markdown 文件。我自己都不读我的 Markdown 文件。那另一个设计师怎么可能知道里面写了什么？你知道，这里有一个可见性的挑战，真的会很有意思。

Yeah, and that's why you see Google still launching the design.md file. You're going to see a lot of initiatives trying to standardize what that means. Like, how do you standardize intent? We're moving in that direction, and it's going to look probably very different a year from now. And how do you share the lessons derived and the best practices derived across an org too? Because something we said in the very beginning is we're not even reading our markdown files. I don't even read my markdown files. So how is another designer going to have any clue what is in there, you know? There's a visibility challenge that's going to be really interesting to think about.

这是个非常好的观点。我之前没想到过。随着我们展开讨论，围绕这些黑盒进行协作，将会是一个需要解决的挑战。完全随机的问题，没错，在你走之前来个出其不意：如果明天你必须在某家公司做一个 W-2 雇员，你想到哪里工作？为什么？

That's a really good point. I haven't thought about that. As we elaborate, collaborating around those black boxes is going to be a challenge that needs to get solved. Totally random question, yeah, left field before I let you go: if you had to take a W-2 role at a company tomorrow, where would you want to work, and why?

天哪，我会选 Vercel，伙计。没错。

Oh man, I think it would be Vercel, dude. Yeah.

真的吗？

Really?

我喜欢他们不端着……他们真的在朝这个方向做事。而我只是一个创造者。我真的相信未来属于创造者。前几天我说过一句话：我觉得未来是光明的，但必须靠人去建设。我喜欢身边都是那种不停建造的人，痴迷于朝着能带来更好变化的方向去建设。那是我见过最酷的文化之一。现在有一种向更技术型产品靠拢的引力，我认为它会吸引到最顶尖的设计师——那些偏创造者类型、更像是早期采用者的人，你知道吧？所以这些文化能吸引到那么高水平的人才，我一点也不意外。因为现在每一个因为突然能实现自己想法而激动不已的人，都想要进入那种被允许放手狂奔的环境。所以，是的，或许组织层级的不同也会让设计师的可能性有一些分化。所以我理解。

I like that they're not full of... Like they're really working in this way, and I am just a builder. I really think the future belongs to builders. There's something I said the other day, which was: I think the future is bright, but it has to be built. And I love being surrounded by people who are just building, who are obsessively building in a direction that betters things. That was one of the coolest cultures I've seen. There's a gravitational pull towards more technical products right now, and I think it gets the cream of the crop of the designers that skew builders—more like the early adopters, you know? So it's no surprise to me that those cultures are attracting the level of talent they are. Because anybody freaking out right now that they can suddenly build their ideas wants to be in that kind of environment where they're told to just run crazy. And so yeah, there is a little bit more of a separation, maybe, in terms of the different tiers of orgs and what that makes possible for a designer. So I get it.

在我放你走之前，我们还有没有什么没聊到的，你……快速植入一下：我们有个派对要来了。

Anything that we haven't talked about before I let you go that you're... Quick plug: we got a party coming up.

我们确实有个派对要来了。如果你在听这期节目，来旧金山参加 Config 吧。最后一晚是 Detach。会特别好玩。Detach 是 Jesse Showalter 和我在两届 Config 之前想到的主意。我记得那也是我们俩认识的地方。我记得当时大概 150 人。就是：会议结束后，我们别搞什么迷你会议了。我们就……我跟 Jesse 说：“我就想和那些我在网上认识的设计师们一起喝酒、一起玩。”然后我们做到了。结果非常棒。去年规模大了一点，我记得大概 300 人。我们每年都卖光票。所以今年我就想，好啊，我们在 25 Lusk 租了三层楼。场地特别酷，还有一个很漂亮的屋顶。我们大概有 400 人。那些来找我的人，比如 Vercel 的团队会来。Designjoy 的 Brett，他从来不出门——他连家门都不出——他说：“我会去的。”我就说：“太棒了。”还有 Josh Puckett。就是有好多好酷的人。我跟你说，如果你对那种真诚、不装腔作势的社交感兴趣，就来参加派对吧。来玩。我想和你喝一杯。你会尝到一点这个社区有多棒。

We do have a party coming up. If you're listening to this, come to San Francisco for Config. And the last night is Detach. It's going to be so fun. Detach is this idea that Jesse Showalter and I had two Configs ago. It's where you and I met, I think. I think it was like 150 people. It was just: once the conference is done, let's not have a mini conference. Let's just... I told Jesse, 'I just want to get drunk with all the designers that I see online and hang out.' And we did. It turned out great. Last year it was a little bit bigger—I think we had like 300. And we've been selling out every year. So this year I'm like, cool, we have a three-story at 25 Lusk. Really cool venue. It's got a beautiful rooftop. And we've got like 400 people. And the people who are hitting me up, like Vercel's team is coming. Brett from Designjoy, who never leaves—he never leaves his house—he's like, 'I'll be there.' I'm like, 'Heck yeah.' Josh Puckett. It's just so many cool people. And I'm telling you, if you have an interest in really networking in a way that's not pretentious or anything, come to the party. Come hang out. I want to have a drink with you. You get a little taste of how amazing this community is.

Tommy，你在其中贡献和塑造的影响力真的相当大。非常感谢你来做客，分享一些你现在的想法和你看到的东西。我觉得我们应该把这些做成一个固定环节，因为你的视角非常好，而且事情变化太快了，不该隔一整年才录一期。

And Tommy, you've played a pretty outsized role in contributing to it and shaping it. Super appreciate you coming on and sharing a little bit about what's on your mind and what you're seeing today. I think we should make a regular practice out of these because that vantage point is a good one, and things are moving way too quickly to have a full year in between episodes.

我们应该有一个半年一度的“视角国情咨文”，你知道吗？

We should have a bi-annual state of the state of perspective, you know?

我觉得这是第一次做这个，以后还会有更多。

I think this is the first one of those, and we'll have more.

顺便说一句，是受你启发——一切都源于你。我懂你。好吧，不过我懂你。

Inspired by you, by the way. That's where it all came from. I got you. All right, but I got you.