AI 时代的设计工程：与 Flora Guo 的对话

我们要如何用 AI 来真正加速我们作为学习者、思考者和设计师的成长过程？

How can we use AI to just really supercharge our own process as learners and thinkers and designers?

AI 最擅长的事情，是给我们一张领域地图。就像玩电子游戏时，地图上没探索过的地方笼罩着迷雾。AI 非常擅长吹散这些迷雾。基本上，你能清晰地看到什么是可能的，但真正走一遍这片土地，仍然要靠我们自己。

What AI is really great at is giving us a map of the territory. It's like when you're playing a video game and you see clouds surrounding areas of the map that you haven't looked at. AI is really good at blowing those clouds away. Basically, now you have a very clear idea of what is possible, but it's still up to us to walk the terrain.

欢迎收听 Dive Club，我是 Rid，在这里，设计师永不停下学习的脚步。今天的嘉宾是 Flora Guo，她的职业生涯轨迹是我采访过的所有人中最亮眼的之一。她目前在 Paradigm 担任创始设计工程师；在此之前，Vercel 的 CEO Guillermo 亲自私信邀请她加入团队。所以，我们将会深入聊聊职业成长、设计工程、在一家 AI 原生的创业公司做创始设计师是什么体验，以及更多话题。但我想用一点不一样的方式来开启这期节目：我请 Flora 分享她在最近的东京设计大会上的几个最喜欢的收获。

Welcome to Dive Club. My name is Rid, and this is where designers never stop learning. Today's episode is with Flora Guo, who's had one of the more impressive career trajectories of anybody that I've interviewed on this show. She's currently the founding design engineer at Paradigm, and before that, Guillermo, the Vercel CEO, personally DM'd her to get her to join the team. So we're going to do a deep dive into career growth, design engineering, what it's like being the founding designer at an AI-native startup, and a lot more. But I wanted to start this episode a little differently. So I asked Flora to share some of her favorite takeaways from the recent Tokyo Design Conference.

Brian Collins 基本上把业界最有趣的 50 位设计师聚到了一起，像 Rio、Brian Lovin、Gabriel Valdivia——很多都上过你的节目。那两天能向所有人学习，真的是无比难得的机会。我是一个随时随地都在记笔记的人，无论是在网站上在线记录，还是用笔记本手写，我真的很想给这些笔记找一个安身之处。最近，多亏一些很棒的合作伙伴，邮票这个想法一直在我脑海里挥之不去。我还有一个朋友 Josephine Ong 也给了我很大启发，她刚上线了一个很酷的个人作品集，用的是无限网格交互——和 It's Not About Cosmos、public.work 上那种类似。我当时就知道，我必须坐下来亲手做一个东西，来满足一下那份创作的渴望。所以，这就是我最终要做的。

Brian Collins basically brought 50 of the most interesting designers in the space, so people like Rio, Brian Lovin, Gabriel Valdivia — a lot of your guests. And it was just such an incredible opportunity to learn from everyone across those two days. I'm someone who is always taking notes, whether it's online on my website or physically in my notebook, and I just really wanted to create a place for these notes to live. And recently, stamps have been living in my head rent-free, thanks to some incredible partners. And I was also really inspired by one of my friends, Josephine Ong, who just shipped such a cool portfolio with this infinite grid interaction — the same one that you'd see maybe on It's Not About Cosmos, like public.work. And I just knew I had to sit down and build something out myself just to scratch that creative itch. So, this is ultimately what I'm going to do with it.

我本人就是那个很想去、但没能去成、懊恼到不行的人。然后你把这些笔记整理发布出来，我心想：“哦，太感谢了。”

I was totally someone who wanted to go, couldn't make it, had incredible FOMO. And then you released all of your notes like this, and I was like, “Oh, thank you.”

我为每一场做了笔记的演讲都制作了一枚小邮票。其中三场让我特别有共鸣——如果我切换到可以把它们都看作页面的视图的话——第一是 Brian Lovin 关于用 AI 构建软件的精彩分享。听他讲 AI 如何通过 ProtoPie Playground 赋能他在 Notion 的团队，以及我们整个行业正在经历的转变，真的非常震撼：我们正在从创造仅供展示的漂亮图片和作品，转向设计师能够展示交互、真正深入细节，去思考产品会如何呈现、最终用户会如何体验。

I made these little stamps for each talk that I had notes for. And three that really resonated with me are — if I go into this view where I can see them all as pages — Brian Lovin did this incredible session about building software with AI. I think it was just so incredible to hear about how AI has been empowering his team at Notion with ProtoPie Playground, and just the way that we're moving as an industry. It's going from creating pretty images and artifacts that are just for show, to designers being able to showcase interactions and really get into the nitty-gritty of how things will display and be experienced by the end user.

另外一场特别突出的是 Rio 关于构建下一代 Cursor 的演讲。你知道，Cursor 是我们大家都在用的工具，但能看到他和小伙伴努力追求背后的那些理念，真的很棒。比如“黑盒 AI 界面”与“玻璃界面”的想法：我们怎样才能真正呈现用户的意图？另外，当我们回顾软件历史，像 Bill Atkinson 和 Alan Kay 这些人，在非常早期的时代，创造设计的人和用代码把它实现出来的人是同一个人。正因如此，最初的意图与用户最终看到、体验到的产品之间，才会有那样紧密的融合。

One that really stood out was Rio's talk on building the next Cursor. You know, Cursor is a tool that we all use, but just getting to see the philosophies that underlie what he's striving for with the team. So, this idea of black-box AI interfaces versus glass interfaces. How can we really surface the intent that the user has? Also, when we look back at the history of software with folks like Bill Atkinson and Alan Kay, in the very early days, the people who were creating the designs and bringing it to life in code were one and the same. And because of that, we just had such a tight integration of their original intention and what people actually got to see and experience.

插播一条简短消息，然后我们继续。过去一年里，Dive Club 让我非常清楚地看到一件事：设计的实践正在改变，而过去那种获取反馈的老办法，在今天的世界已经不太够用了。所以我很兴奋地宣布，Inflight 正式进入公开测试阶段。它是我一直想要的反馈工具，专为一个以 AI 速度运转的世界而打造。我可以分享原型、用视频演示来提供背景，Inflight 能轻松帮我获得前进所需的精准反馈——无论是给方向投票，还是给新想法开绿灯放行。而且所有这些，都可以通过一条链接完成，我可以丢到 Slack 里，或者分享给核心用户来测试新原型。我每天都在用 Inflight，它彻底改变了我分享工作的方式。我很期待你来试用这个产品；如果你想聊聊它，给我发邮件：rid@inflight.co。

Real quick message, and then we can jump back into it. 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.

如果你一直在听这档节目，你应该已经听我多次表达对 Raycast 的热爱了。所以我想特别提几位在 Dive Club 上听说 Raycast 之后开始使用它的朋友。第一位是 Rex Harris，他说 Raycast 是塑造他使用电脑方式、让他保持心流状态的最重要工具。接下来是 Greg Hunton，他称这是一个改变人生的产品，说每当他顺着某个线索深入 Raycast 时，使用电脑的体验和力量都会被大幅提升。我完全同意，而且怎么推荐 Raycast 都不为过。入门完全免费，前往 dive.club/raycast 即可查看。

If you've been listening to this show for a bit, then you've heard me profess my love for Raycast multiple times by now. So I want to highlight a couple of people who started using Raycast after hearing about it on Dive Club. The first is Rex Harris. He says Raycast is the tool that most shapes how he uses his computer and keeps him in flow. The next is Greg Hunton. He called it a life-changing product, and says every time he pulls a thread on something in Raycast, it wildly deepens the power and experience of using his computer. I couldn't agree more, and I can't recommend Raycast enough. It's totally free to start. Just head to dive.club/raycast to check it out.

好，回到正题。有一场演讲真的让我有强烈的“错过一个亿”的感觉，就是 Saelio 的《好运的几何学》（Geometry of Luck），他在 Twitter 上预热的时候，我恨不得自己就在现场。所以，你听那场演讲的体验如何？能不能分享一下你的笔记？

Now, onto the episode. So, there's one talk that really brought the FOMO for me, which is Saelio's Geometry of Luck, and he was teasing it on Twitter and I just wish I was in the room. So what was your experience with that one, and maybe could you share a little bit of your notes there?

我觉得这绝对是一场让我非常有共鸣的演讲，因为 Saelio 提出，运气并不是——你知道的——你与生俱来的东西，或者天上掉下来的馅饼，而是有特定的形状和几何结构的。这意味着，就像任何其他技能一样，它是我们可以通过学习去了解、并邀请更多进入我们生活的东西。他带我们看了几个不同的例子，比如 John Chapman，或者说 Johnny Appleseed。从旁观者的角度看，你可能会想：“哇，他在移民进入美国之前就撒下这些苹果种子、种下这些果园，真是太有先见之明了。”但实际上，他只是研究过地图——比如河流流向哪里、人们在什么季节会来，然后对周围发生的一切保持高度的觉察。就像我们一样：我们看到 X 上不断涌来的海量内容，隐约感知到各种可能性，由此让自己处于一个有利的位置，能够真正投入进去，开始用这些工具去探索。那些认为自己幸运的人，也能够在机会到来时识别出机会。他们为自己创造了更大的“运气接触面”，无论是身处某些圈子，还是去创造那些能够产生意外之喜的圈子。我觉得，往往就是一个瞬间，就能彻底改变你的人生轨迹。

I think this is definitely a talk that really resonated with me, because Saelio proposed that luck isn't, you know, something that you're born with or just falls in your lap, but it has a specific shape and a geometry. And what that means is that, like any other skill, it's something that we can just learn about and invite more of into our lives. So he walked us through a few different examples — people like John Chapman, or Johnny Appleseed. So, someone who from the outside looking in, you're like, “Oh wow, he had so much prescience in scattering these apple seeds and planting these orchards before settlers came into the States.” But he was actually someone who just studied the maps of — you know — where are the rivers going, what seasons will people be coming in, and then just being so aware of everything that was happening around him. Kind of like for us: we see this constant deluge of content coming in from X, just being dimly aware of what the possibilities are, so that you're well positioned as someone to really jump in and start exploring with these tools. People who just perceive themselves to be lucky are also able to identify the opportunities as they come to them. They're creating a greater surface area for this luck, whether it's being in rooms or creating those rooms where serendipity can happen. I think it's often times just like one moment that can completely change your trajectory.

这个项目本身就是那个想法的萌芽——你把自己喜欢的东西分享出来，放到网上。我不记得它在 Twitter 上获得了多少展示量，但一定很多，差不多有 10 万。显然大家很欣赏你的作品，引起了共鸣。这让我想知道，你在多大程度上有意识地经营自己面向好运的“表面积”。也许你可以谈谈，在线分享在你自己的职业生涯中扮演了怎样的角色。

This project in itself is almost an inception of that very idea because you've shared this thing that you enjoyed and put it out on the internet. I don't remember how many impressions it got on Twitter, but it was a lot — almost 100,000 or something. People obviously enjoyed your work; it resonated with them. And that makes me wonder how intentionally you are stewarding your own surface area for luck. Maybe you could talk a little bit about the role that sharing online has played in your own career journey.

我经常引用一句话，出自雷·伊姆斯或查尔斯·伊姆斯：设计师的角色就像一位很好的主人。在今天这个世界里——就像你说的，竞争极度激烈，新工具层出不穷，总有很多事情在发生——我觉得真正要问的是：我们能给予什么？无论是为一个活动留出空间，还是围绕我们热爱并想要分享的事情创造一件作品。这其实是把问题反转过来了：不是“我能向这个人索取什么”，而是“我能为他们提供什么”。这是我在线上或线下出场时，不断想回到的心态，尤其是我们举办活动的时候：邀请人们进入一个空间，想想要把他们聚拢在什么主题周围？我想为这件事设定什么样的意图？我们在线上经营空间的方式也很类似：无论是把人们聚在一起看一部设计影片，还是邀请他们展示自己的作品。我觉得关键在于，找出把大家联结在一起的那条线索——我们为什么要在同一个空间里——然后再去想，我怎样也能把这种联结从别人身上激发出来。我真的非常感激那些在网上遇到的人。一开始也许只是看到彼此的作品，心想“哇，这也太酷了”。然后你发一条私信，接下来可能就是线下的咖啡或电话聊天，再然后就会擦出合作的火花，无论是项目协作还是工作机会。我做了几年独立设计师，可以说绝大多数客户都来自 Twitter。作为一个一直喜欢实验、喜欢凭心情尝试新事物的人，这真是认识志同道合的人的好方式。就连我现在的这份工作也充满了机缘巧合——我的 CEO Anna 是在 Twitter 上发现我的。在 Bracel 的时候也是一样，一切源于私信里的对话。所以我觉得，任何正在做项目的设计师都活在这样一个不可思议的时代：可以把作品分享给整个互联网。

There's this quote I often return to, by I think either Ray or Charles Eames, that the role of a designer is that of a very good host. And I think showing up in today's world — which, as you said, is ultra-competitive, with a lot going on and all these new tools — is really about asking what we can give, whether it's holding space for an event or creating an artifact about something we're excited about and want to share. It's really flipping the script: instead of 'what can I ask of this person,' it's 'what can I offer them?' That's something I want to keep returning to in how we show up online or in person, especially when we host events — inviting people into a space, thinking about what I can gather them around, what intention I can set around this. The ways we host spaces online are similar, whether it's gathering people around a common interest, like screening a design film, or inviting them to show off their own work. I think it's really about figuring out what threads connect us all together — what is the reason for us to be together in this space — and then figuring out how I can pull that out of others as well. I'm incredibly grateful to the people I've gotten to meet online. People start off maybe just seeing each other's projects and thinking, 'Wow, this is really cool.' Then you send a DM, and it kicks off a conversation IRL over coffee or a phone call, and then it sparks either a collaboration on projects or even roles. When I was working as an independent designer for a couple of years, I'd say the vast majority of my clients came from Twitter. As someone who's always loved experimenting, trying things out for the heck of it, it's been such a great way to meet people with similar interests. My current role felt very serendipitous — my CEO, Anna, found me online on Twitter. And when I was at Bracel, it was also through conversations that started in the DMs. So I think for any designer building projects, it's such an incredible time to live and get to share with the internet.

你能再深入谈谈，你是如何有意识地培养自己在 Twitter 上建立的那个社群的吗？我觉得很多人看到你现在的状态，会觉得非常励志，但同时也像谜一样，不知道怎么才能走到那一步——让 Twitter 突然成为职业生涯中非常重要的一环。

Can you go a little deeper into how intentionally you are nurturing the community you've built for yourself on Twitter? I think a lot of people look at where you are now, and it's really inspiring, but also a bit of a black box in terms of how to even get to that point where Twitter becomes a very big deal for your career.

我记得那是在 2022 年 3 月——我第一次上 Twitter。那时我只是分享自己特别兴奋的东西：去参加线下活动，然后写“哇，我刚学到了设计里的某个东西，这是我的收获”。仅仅是这样出现、带着热情、真心想聊天，事情就一环接一环地发生了。有些角色最初只是来自一条私信，后来发展成线下的合作，比如一起办活动，甚至最终共事。人们真的能感受到你的兴奋，无论是通过一个项目、一些笔记，还是你主办的某件事。所以关键在于，创造一些能承载你热爱之物的“作品”，然后把它们放到网上。

I remember back in March 2022 — that's when I first got on Twitter. It was just me sharing things I was really excited about: going to events IRL, posting things like, 'Oh wow, I just learned about X in design, here are my takeaways.' Just showing up, being excited, and really down for conversations led one thing to another. Roles that started from a DM evolved into in-person things, like hosting an event together, or eventually working together. People can really feel your excitement, whether through a project, through notes, or through something you're hosting. So it's about creating things that are artifacts of what you love and are passionate about, and putting them online.

一种方式是把自己的作品放到网上，吸引那些真正在乎的人；另一种方式是主动私信别人，成为主动发起交流的那一方。后者你做了多少？我觉得这非常宝贵，但对很多听众来说可能有点吓人。

There's putting things out there on the internet and being a magnet for people who care. And then there's also proactively messaging others and being the one who initiates. How much are you doing the latter? I think that's something so valuable, and yet probably a little bit scary for a lot of people listening.

每当我看到一位我非常欣赏其作品的设计师，我就喜欢给他们发一条简短的私信，比如：“嘿，我喜欢你发布的东西，尤其是这里，真的触动了我。”我每次收到这样的私信，都会开心一整天。我觉得办活动也是接触别人、也让别人接触你的好方式。无论是设计电影之夜、设计工程 PowerPoint 之夜，还是关于智能体式界面的活动——关键是创造一些非常具体、可以提供给别人的东西。比如，“来我家或我办公室吧，我很想招待你。”或者，寄给他们一个你觉得他们会喜欢的东西。这样就创造了一些更自然的时刻——就像你在网上和一个新朋友聊天。

Whenever there's a designer whose work I really admire, I love dropping a quick note like, 'Hey, I love what you shipped, and here in particular is what really resonated with me.' Whenever I get those in my inbox, it really makes my day. I'd say hosting events is also a great way to reach out to people and have them reach out to you. Whether it's a design film night, a design engineering PowerPoint night, or something about agentic interfaces, I think it's about creating very specific things you can offer people — like, 'Come into my home or my office, I'd love to host you.' Or sending them something you think they'd enjoy. It creates moments where it feels much more natural — you're just talking to a new friend on the internet.

我在这一连串话题上问了这么多后续问题，是因为我看到你的轨迹陡峭得惊人——你能多快融入这个领域并建立起势头。现在你在 Paradigm 担任这么受瞩目的角色，真的很酷。我这次对话的一个目标，是帮大家找到那个等式里的小碎片，让他们可以用来反向拆解你的轨迹。所以也许我们可以再多聊聊这段旅程，从 Vercel 的设计工程实习开始。

I think the reason I had so many follow-up questions on this category of topics is that I look at your journey and see an incredibly steep trajectory — how fast you've been able to insert yourself and build momentum in this space. Now you're at Paradigm in this high-profile role, and it's really cool. A goal of mine in this conversation is to help people find little pieces of that equation so they can reverse engineer your trajectory. So maybe we could talk a little more about what that journey looked like, starting with the design engineering internship at Vercel.

Vercel 是我仰慕很多很多年的公司——甚至在我还知道什么叫部署之前就是如此。我记得第一次访问他们首页时心想：“哇，这网站做得也太美了。”我其实是从 v0 开始我的设计工程之旅的。我记得 2024 年的时候，我在网上四处看，时间线上冒出那么多令人惊叹的项目，完全不知道从哪入手。这些工具不可思议的地方在于，你带着一个想法就能开始，慢慢摸索出通往一个让自己骄傲的作品的路径。所以我最初是做一些着色器——看起来像日出的那种——然后也在 v0 里做个人项目。其中我做得特别开心的是一个“数字礼盒”的想法。那是我第一个坐下来对自己说“好，今天我要做点东西”的项目。我真的很喜欢“软件作为礼物”这个概念。

Vercel is a company I've looked up to for many, many years — even before I knew what a deployment was. I remember being on their homepage and thinking, 'Wow, this is just so beautifully built out.' I actually started my design engineering journey with v0. Back in 2024, I was looking around and seeing such incredible projects popping up in my timeline, with no idea where to start. The incredible thing about these tools is that you can go in with just an idea and start noodling out what the path looks like to creating something you're really proud of. So I started off by making shaders — things that looked like a sunrise — and then moved on to personal projects in v0 as well. One that I really had a lot of fun making was this idea of a digital gift box. That was the first project where I just sat down and thought, 'Okay, today I'm going to make something.' I really love the idea of software as a gift.

我觉得做这些项目一直是我表达那种意图的好方式——在互联网上创造一个空间，把想分享的东西打包起来，比如图片、链接，然后发给朋友。这样的分享最后被 Vercel 团队看到了，我们也会进行一些零散的对话。大约一个月后，也就是 1 月，G 在 Twitter 上私信我：“你没想过到 Vercel 工作吗？”

I think these are always an incredible way for me to translate that intention of creating a space on the internet where I could package things up for my friends, whether images or links, and send them over. Sharing it like that basically found its way to the Vercel team, and we would have these one-off conversations. Eventually, about a month later, in January, G DM'd me on Twitter and was like, 'Haven't you ever thought about working at Vercel?'

不会吧。

No way.

那真的太疯狂了。我以前确实给他发过消息，我记得只是说：“嘿，真的很感激你。谢谢你做了 Vercel。我用得很开心。”

And that was really crazy. I had messaged him in the past, I think, just saying, 'Hey, I really appreciate you. Thank you for building Vercel. I'm having so much fun using it.'

好吧，那真的太棒了。我要对所有听众强调一下。你给 Vercel 的 CEO 发私信，只是为了说声谢谢，这充分说明了你的为人，而且这也是我前面想说的，对吧？这很了不起。然后，你知道后来发生了什么吗？我打赌他忘了。他绝不可能真的记得。绝不可能。但后来他看到了那个项目，也可能是有人转发给了他。他点进去，心想：“好，也许我可以联系一下这个人。”而那个页面顶部就是你的感谢留言。他大概会说：“哦，哇，现在我真的很想和这个人聊聊。”所以这太棒了。我只是想把这件事拎出来说说。我觉得你在这个社区里的自我定位真的很棒。另外，G 也是个在网上超级主动联系别人的人。我们在 Vercel 的很多员工，都是他在网上看到后说“我们得和这个人聊聊”。他有个叫“G Brain”的笔记系统，会在里面记各种小笔记，然后过段时间就想：“好，招聘这边跟一下。”正因为他就是这样的人，他在 Vercel 乃至整个世界都创造了很多机缘巧合。我相信你当时肯定特别激动。你收到这条私信，感觉不可置信，对吧？你最后加入了团队。你在职业生涯早期就与全世界最优秀的创造者在一起。我猜你当时也在想，自己到底要成为这个领域里的谁。那么，这段经历如何塑造了你对想成为的那类设计师的心智模型？

Okay, that's really cool. I'm going to underline that for everyone listening. The fact that you DM'd the CEO of Vercel just to say thank you says so much about you, and it's kind of what I was getting at earlier, right? That's a big deal. And then, you know what probably happened? I bet he forgot. There's no way he actually remembered. There's no way. But then he saw the project, or maybe somebody sent it to him. He clicked in, said, 'Okay, maybe I'll reach out to this person.' And at the top of that page was your thank-you message. And it was like, 'Oh, wow. Now I really want to talk to that person.' So that's amazing. I'm just shining a light on it. I think it's amazing the way you position yourself in this community. I think G is also someone who's incredibly proactive about reaching out to people online. So many of our hires at Vercel were people he saw on the internet, and he was like, 'We need to talk to this person.' He has this mental note system called 'G Brain' where he puts all these little notes, and then it's like, 'Okay, now what's recruiting? Let's follow up.' Just being the type of person he is, he creates so much serendipity both at Vercel and in the world. I'm sure you were freaking out. You get this DM. It's unbelievable, right? You end up joining the team. You spend this chunk of time earlier in your career surrounded by some of the best makers on the planet. And my assumption is you were probably thinking a little bit about who you even want to become in this space. So, how did that experience shape your mental model for the type of designer you wanted to grow into?

能与 James Clements、Matias Shaw 和 Rono 这样的人共事，真是一种超现实的体验。看到他们对自己作品的专注程度，无论是打磨营销网站的交互，还是深入钻研应用的性能，你都会意识到：设计工程远比事物表面的样子要深。它关乎拥有上下文，并且有能力把一个意图转化为用户真正体验到的内容。在 Paradigm 也是如此，我们一直在思考：怎么保证我们脑海中的愿景，就是用户在生产环境里真正玩到的东西。而且那些总是学习新技术、新工具的设计师身上，有太多值得学习的地方。他们就是出于好奇心去学。Rono 让我印象特别深的一点是，他总是在做各种副项目，不管是开发细节还是个人网站。被一群为了热爱而创作的人包围，真的很受鼓舞——他们不是因为要完成某个项目，也不是因为把事情做完就是终点。他们真正把设计看作一段永无止境的实践旅程。在这条路上行走，本身就有很多美好。

It was such a surreal experience to work alongside people like James Clements, Matias Shaw, and Rono. Just seeing the level of care they put into their work, whether it's crafting interactions for a marketing site or really diving into the performance of the app, really shows you that design engineering goes much deeper than the surface of how something looks. It's about having the context and the ability to take an intention and translate it into what the user is actually experiencing. At Paradigm, it's also something we're always thinking about: how can we make sure that the vision we have is what people actually get to play with in prod? And there's so much to learn from the designers who are always picking up new technologies and tools, people who are just doing it because they're really curious. Something that really stood out to me about Rono is how he's always cooking on side projects, whether it's developing details or his personal websites. It's so inspiring to be surrounded by people who make for the love of making, not because you're looking to wrap up a specific project or just get something done because that's the finish line. It's really appreciating design as this journey, this practice that never ends. And there's a lot of beauty in walking that path.

我前几天随便点进了 Rono 的网站，它又变了——整个风格完全不一样。你知道设计师圈子里有个笑话：我们永远找不到时间更新作品集。但这个家伙每隔 6 个月就会有一个做工精致的新个人网站。你能看得出来，对他来说重点不是最终产出，而是创作本身。我不知道怎么形容，但那一直是我作为设计师想成为的样子。就是做一个因为创作能滋养灵魂而去创作的人。我觉得尤其是当我们在尝试新东西的时候，看到网上有这样一些可以仰望的人——‘哇，他们做了个超酷的项目，在这里讲了是怎么做的、用了什么技术栈、在玩哪些工具’——这就会推我们一把，让我们想：‘好，今天我要坐下来搞清楚这个 paper MCV 到底怎么运作。’或者：‘今天我要坐下来把我的 Claude Code 文件配置好。’有这样的榜样，对我来说真的非常有力量。

I was clicking through Rono's website randomly the other day, and it was different again—just a completely different everything. And you know the joke as designers: we can never find the time to update a portfolio. And this dude has a new personal website dripping in craft every six months. You can tell it's not even about the end output for him. It's the act of creating, which I don't know, has always been something I aspire to grow into as a designer. It's just being somebody who creates because it's life-giving for my very soul. I think especially when we're testing things out, having people online we can look to—'Oh wow, they built this really cool project. They're talking about how they built it, the stack they used, the tools they're playing with'—it just gives us that extra nudge to be like, 'Okay, today I'm going to sit down and figure out how exactly this paper MCV works.' Or, 'Today I'm going to sit down and set up my Claude Code files.' Just having role models like that has been incredibly powerful for me.

在进入下一个话题之前，给我们讲几个你最喜欢的实验项目吧，作为‘纯粹为了创作而创作并发布到网上’的例子——因为我觉得你的 Twitter 时间线很漂亮。所以我暂时不想往下进行。

Before we move on, take us through a few of your favorite experiments as an example of what it looks like to just create for the sake of creating and put it out there on the internet—because I think your Twitter feed is beautiful. So I don't want to move on quite yet.

有一个我觉得特别傻的项目，是我朋友最喜欢的电影是《超级无敌掌门狗》。我想给他做一个可以在主屏幕上玩的小东西，也就是他最喜欢的角色“羽毛麦格劳”，还有一些剧情细节。这更像一个你可以挂载的视觉小物件，比如建一个 3D 模型，再做一个 3D 模型查看器，然后把一些东西作为礼物送给人。我觉得这一点你们会和我有共鸣。还有一个是很久以前的项目——我刚入门做项目的时候，常常不知道怎么入手。于是我做了个非常简单的网页应用：你提出任何想法，它会帮你拆解，比如前端可能要思考哪些问题，后端可以学些什么。对我来说，它更像一个“地图生成器”。这个项目我真的很喜欢。它又回到了“软件即礼物”这个想法——我们能用什么方式创造一些有触感的东西，创造一些你可以发给别人，然后说“嘿，这让我想到了你”的东西。最近我也很喜欢玩动画，比如 Product 或 Tokyo 项目，就是在琢磨可以玩哪些酷炫效果。这个项目我必须感谢 Daniel Pito，他发布了一个 CSS 小片段，教你怎么做一个 3D 无限画布，让东西向你飞来。

One that I thought was really silly was my friend's favorite movie is Wallace and Gromit. I wanted to make him a little artifact he could play with on his home screen—his favorite character, Feathers McGraw—with little plot details. It's more of a visual artifact you can hook up, create a 3D model, and then create a 3D model viewer, and just give something to people as a gift. I think that's something you guys can resonate with. This is a project from way back when I was starting off with making projects and wouldn't really know how to approach it. So I made a very simple web app where you take any idea and it breaks it down for you: here are things you might want to think about on the front end, here are things you can learn about for the back end. It served as more of a map-maker for me. This was a project I really enjoyed. It returns to the idea of software as a gift—what are ways we can create things that feel tactile, things you can send to people and just say, 'Hey, this made me think of you.' Recently, what I really love as well is finding ways to play with things like animations, whether it's the Product or Tokyo project. It was really about figuring out cool effects I can play with. This one I really have to shout out to Daniel Pito: he put out this CSS snippet showing how to create a 3D infinite canvas where things come at you.

当我在摆弄 3D 画布时，我其实是在实验性能。Maxim Petkov 是我的一位朋友，他是位非常出色的设计工程师。我把东西发给他，他说：‘哦，顺便说一句，如果你把 CSS 类改成 transform，性能会好很多。’所以我一直在玩一些一直放在我缓存里的新东西，然后把它们转化成我实际在做的项目。

And when I was playing with the 3D canvas, I was just experimenting with performance. Maxim Petkov is a friend who's an incredible design engineer. I sent it over to him, and he said, 'Oh, by the way, if you change the CSS class to transform, it'll be a lot more performant.' So I'm always playing around with new things that have been sitting in my cache and translating them into a project I'm actually working on.

很明显你在前端方面学到了很多。你越是去构建自己的想法，我就看到越多小技术的例子——比如你为《超级无敌掌门狗》做的那个 3D 渲染。我完全不知道那是怎么实现的，但你在每个项目里把它弄明白了，很明显这些都在帮你爬上下一级学习阶梯。这让我很好奇：这一切的起点在哪里？你第一次打开 V0 时有多少技术理解？也许我们可以以此为出发点，了解你一路走来的那些理解里程碑。

It's clear that you're learning a ton about front end. The more you build your ideas, I see a lot of examples of little technologies—like you had that 3D render for Wallace and Gromit. I have no idea how that works, but you're figuring it out for each individual project, and it's obvious those are helping you reach the next ladder of learning. That makes me wonder: where was the starting point for all of this? How much technical understanding did you have when you first opened V0? Maybe we can use that as a launching point to understand some of the mile markers of understanding for you.

我记得刚开始玩 V0 的时候，我对 HTML、CSS 和 JavaScript 只有非常基础的掌握。我也是一个很喜欢上网课的人。其中一门课是 Mariana Castillo 在 General Assembly 开设的 UI 工程 101。

I remember when I first started to play around with V0, I had a very basic grasp of HTML, CSS, and JavaScript. I'm also someone who loves picking up internet courses. One was UI Engineering 101 by Mariana Castillo with General Assembly.

哦，真棒。

Oh, cool.

是的，我学完了那门课。我很喜欢他们带你一步一步走：这是一个组件，这是如何仅用 HTML、CSS 和一点 JavaScript 来实现它。后来我读了很多 React 文档，试图弄明白 React 框架是什么、Next.js 是什么。我的想法是：总有那么一点点最小可行的学习量，能让你走到下一步，然后再补很多课。AI 非常擅长给我们产出制品，但这些制品不一定建立在对系统的理解之上。这在做原型和小的一次性实验时很棒，但当你要调试某个东西时，你会想：‘天哪，我甚至不知道从哪里开始。’

Yeah, I went through that. I loved how they walked me through: here's a component, here's how you translate it with just HTML, CSS, and a little bit of JavaScript. Eventually, along the way, I was reading a lot of React documentation, trying to figure out what React frameworks are, what Next.js is. The way I think about it is that there's always a minimum viable amount of learning that gets you to the next step, and then you backfill a lot. AI is really good at giving us artifacts, but they're not necessarily grounded in understanding of the system. That can be great for making prototypes and little one-off experiments, but the moment you need to debug something, you're like, 'Oh, man, I don't even know where to start.'

嗯。

Mhm.

这时候，推自己一把、去感受自己对这些工具还没有完全掌握，然后退一步问‘好吧，我现在需要放大到哪个层面，真正加深对哪里的理解？’会特别有帮助。对我来说，当我开始玩这些不同工具时，我就是没搞懂组件应该怎么组织——什么时候该拆成模块，什么时候该传 props，而不是共享 context。通过构建来学习、通过提问来学习，是一种非常棒的方式：看网上的参考材料，问那些可能是优秀工程师或做过类似事情的朋友。

That's where it's really helpful to push yourself to feel like you don't have a complete grasp with these tools, and then step back and ask, 'Okay, what do I need to zoom out on and really deepen my understanding of now?' For me, when I started off playing around with these different tools, I just didn't grasp how components should be structured—when to break things down into modules, or when to pass down props versus sharing context. Learning by building and learning by asking has been incredible: looking at reference material online and asking friends who might be good engineers or people who have done it before.

很多人都会感受到一种张力，它正好触及你在说的一些东西：我到底该在自己脑子里存多少知识？也许其中一些是语法、代码层面的知识，而不是仅仅改善我与 AI 协作的方式。也许我把很多细节抽象掉，然后把所有注意力都放在那上面。在一切都不断变化的情况下，这是一个值得思考的有趣光谱——你如何看待学习，以及你想把精力集中在哪里。你觉得自己在这个光谱上想站在哪个位置？

There's a tension that a lot of people experience that hits on some of the things you're talking about. It's like, how much do I actually invest in my own knowledge that exists in my brain? Maybe some of that is syntax, code-level knowledge, versus just improving the way that I work with AI. Maybe I abstract a lot of the details and focus all my attention on that front. With everything changing, that's an interesting spectrum to think about—how you view learning and where you want to concentrate your efforts. How do you think about where you want to fall on that spectrum?

对我来说，我觉得这是去年我非常切身经历到的事，因为我做了很多个人项目。我会想：‘哦，不错，看，这是一个 demo。’但当我真的要调试某个东西时——比如我在调应用 API，突然它加载得很慢或者所有东西都崩了——我就会觉得自己没有足够的基础知识去真正深入钻研那一层层文档。AI 最擅长的是给我们一张领域地图。就像你玩游戏时，地图上有些区域还被云盖着，你还没解锁。AI 很擅长把那些云吹散。基本上，现在你很清楚什么是可能的，但仍需要我们自己走一遍地形。看到我们能做出什么是一回事，完整地理解它又是另一回事。而且我觉得随着 AI 的发展，我们也在往上移动抽象层。所以我们不会手写每一个函数——没人这么做了。但你对它底层原理了解得越多——React 怎么工作、props 是怎么传递的——我们就越有能力去真正调试并构建有韧性的东西。这取决于我们工作在哪个抽象层。比如，如果一个人在和一支非常强的工程师团队合作，对他们最有帮助的可能是‘这是一个代码原型，正是我想要的’，然后工程师帮你接好整个数据层。也许他们从一个 PR 开始，里面只有所有 API 调用的骨架，以及后端如何与前端通信。然后作为设计师，你可以用 Claude Code 进去打磨一切，确保它在微交互和动画方面完全符合你想要的样子。或者，如果你想在实现层面更深一些——理解这个组件应该如何结构以达到最佳性能——我觉得回到基础、去学习，是我最近更常回到的事情。

For me, I think this is something that I experienced very acutely last year because I was making a bunch of personal projects. I'm like, 'Oh, cool, look, it's a demo.' But when it actually came down to debugging something, if I was playing with an app API and all of a sudden it wasn't loading well or everything was crashing, I just felt like I didn't have the necessary foundational knowledge to actually hone in and dive deep into the levels of documentation. What AI is really great at is giving us a map of the territory. It's like when you're playing a video game and you see clouds surrounding areas of the map that you haven't unlocked yet. AI is really good at blowing those clouds away. Basically, now you have a very clear idea of what is possible, but it's still up to us to walk the terrain. It's one thing to see what we can make and another thing to understand it in its full entirety. And I think with AI we're also moving up the layers of abstraction. So we're not writing every function by hand—no one's doing that. But the more context you have about how it works under the hood—how React works, how props are being passed through—the better equipped we are to actually debug and build things that are resilient. It depends on the level of abstraction that we're working at. If we're someone who's working with a very strong team of engineers, what's most helpful for them is maybe 'here is a prototype in code of exactly how I want it,' and the engineer hooks up all the data layer for you. Maybe they start off with a PR that's just the skeleton of all the API calls and how the back end talks to the front end. Then as a designer, you can go in and polish everything using Claude Code and make sure it's exactly how you want it in terms of micro-interactions and animations. Or if you're someone who wants to be deeper down in terms of implementation—understanding how this component should be structured to be the most performant—I think going back to the basics and studying has been something I've returned to much more lately as well.

有一个问题我一直忍不住问自己：如果公司反过来申请来和你交谈，而不是你去申请，会怎样？这个问题就是全新 Dive 人才网络的基础。而且它正在起作用。现在，我正在帮助我认识的最令人兴奋的许多初创公司，去雇用那些收听本节目的设计师和构建者。所以，如果你想知道外面有什么机会，也许你想上我的名单，或者你甚至正在寻找你的下一个设计人才，请访问 dive.club/talent 立即加入。

There's one question that I can't stop asking myself. What if companies applied to talk to you rather than the other way around? And that question is the foundation for the all-new Dive Talent Network. And it's working. Right now, I'm helping many of the most exciting startups that I know to hire the designers and builders who listen to this show. So, if you're curious what might be out there, maybe you want to get on my list, or maybe you're even looking for your next design hire, head to dive.club/talent to join today.

你是一个认真对待学习的人。光是听你讲，这一点就非常清楚。所以我想问一个稍微深入细节的问题：如果你回顾一下你每天编码时使用 AI 的方式，再看看六个月前的你与今天的你，最大的差异在哪里？

You're someone that takes learning seriously. Just listening to your talk, that is something that's very clear to me. So I want to ask a little bit of an in-the-weeds question here. If you would throw up how you work with AI on a daily basis when coding, and you look at the version of yourself six months ago versus the version today, where do the biggest deltas exist?

我觉得最大的变化之一是，我会一边用 Claude 一边生成解释文档。

I think one of the biggest differences has been creating explainers as I go with Claude.

比如，以前我学习 React 如何与 React Hooks 配合时，我会边写边想，“哦等等，这东西完全可以抽象成一个自定义 Hook。”然后我会翻出一些文档，和 Claude 聊一下，最后又问它：“好，下次要记住哪些要点？哪怕只是对我自己的学习来说。”因为有了技能文件，你知道，你可以给 Claude 上下文。但我也在想，当我们与 AI 协作时，我们建立起来的心智模型，比如 skill.md、fold.md，真的会促使我们去思考：“有哪些方法可以建立我们自己的上下文？”或者：“有哪些方法可以定义我们自己的不变量？”所以我特别喜欢做的是，当我觉得自己尝试了新东西，或者学到了以前不知道的东西时，就会快速写一个 Markdown 文件存下来，留到下次给自己参考。有时我也会把这些笔记发到网站上，但最重要的还是创造一些我可以不断回来查阅的产物。我觉得这是最大的解锁之一。

For example, in the past when I was learning about how React works with React hooks, I would be building it and be like, “Oh wait, this is totally something that can be abstracted out into its own custom hook.” So then I would pull up some documentation, have a conversation with Claude, and then at the end say like, “Okay, what are the key points that we should remember for next time, or even just for my own learning?” Because with skill files, you know, you can give Claude context. But I also think about how, as we work with AI, these mental models that we’re building up, like skill.md, fold.md, really encourage us to think about, “What are ways that we can build up our own context?” Or, “What are ways that we can define our own invariants?” So something I really love doing is, after I feel like I’ve tried something new or I learned something that I didn’t know before, I’ll quickly write up a markdown file and just save that for next time for my own reference. And sometimes I’ll post these notes to my website, but just creating artifacts that I can continually return to. I think that’s one of the biggest unlocks.

你有没有发到网站上的例子，可以让我们看看？

Is there an example that you posted to your website that we could look at?

如果你去看我的笔记，我记得 monorepo（单仓库）是我们在 Vercel 大量使用的东西。从 Vercel 的营销网站到应用，再到那些 Ship Conference 的官网，都托管在一个叫做 front 的庞大仓库里。我记得第一次在里面摸索时，我心里想：“哇，子目录也太多了。”而对于我自己的个人网站，它是托管在 Next.js 应用上的，但我想做一个可以跨项目共享的组件系统。我当时在想：“我该怎么做？Turborepo 在哪里？该怎么设置？”然后就是深入钻研，也顺便写下笔记留着下次用。我觉得这真的很有帮助。

If you look at my notes, I remember monorepos is something we use a lot at Vercel. Everything from the Vercel marketing website to the app to all those Ship Conference websites is hosted in this massive repository called “front”. And I remember poking in for the first time and being like, “Wow, there are so many subdirectories.” And for my own personal website, it’s hosted on a Next.js app, but I wanted to create a shared component system that I could use across projects. I was figuring out, “How do I do that? Where’s Turborepo? How do I set this up?” And then deep-diving into things and just writing down notes for next time as well. I think that’s been really helpful.

我觉得这里面最酷的部分其实是时间戳。比如说你有多频繁地把自己学到的东西以笔记形式发布到自己的网站上。把网站看作活的产物——你知道，就是能够随手记下来。如果你看到一个很酷的链接，或者某个想留到下次再用的东西。我觉得“创造一些你可以反复回去查阅的东西”已经成了非常鲜明的主题。我很喜欢。

I think the part of this that actually is the coolest is just the timestamps. Like how often you are contributing notes from your own learnings to your own website. To think of websites as living artifacts – you know, being able to just jot things down. If there’s a cool link that you saw, something you want to save for next time. I think just creating things that you can return to has been such a theme. I love it.

还有没有其他影响你和 AI 协作方式的心智模型？这些模型是你过去一年左右逐渐打磨出来的，而且你觉得可能会启发其他人？

Any other mental models that influenced the way that you collaborate with AI that you feel like you’ve refined over the last year or so, that you think might inspire other people?

我真的很喜欢把 AI 看作一个无限耐心的导师。你可以让它帮你做事，而不质疑它的工作方式或它基于的假设，这是其中一面。但真正要采取的姿态是：我们现在生活在一个可以问任何问题的时代，而且我可以让它按照我需要的精度，把它拆解到任意细致的程度。我真的鼓励大家去思考，有哪些方法可以用 AI 来搭建自己的学习支架，然后通过做项目去充实它，并试着理解：我们如何提示自己，让自己成为更好的学习者。

I really love thinking about AI as this infinitely patient tutor. I think it’s one thing to ask it to do something for me and not question anything about how it works or what assumptions it’s building off of. But really adopting this posture of, now we live in a time where I can ask any question, and I can ask it to break it down into as fine a level of fidelity as I need. I would really encourage people to just think about ways that they can use AI to scaffold their own learning, and then to flesh it out by building projects, trying to understand what are ways that we can prompt ourselves to become better at learning.

这些想法启发我的一个方式，就是看看你现在的角色。比如，你是一名设计工程师，但——如果我说错了请纠正——Paradigm 并没有传统意义上的设计师，对吧？你某种程度上就是设计师，但你同时也承担了大量前端工作。这条路在理论上，很多听这期节目的人都能走。我其实觉得我自己也正走在这条路上。也许在有些工程概念上没走得那么快，但我是设计师，而且我越来越多地承担前端。对于历史上对代码有些畏惧的人来说，这可能会让人害怕。但这份蓝图，以及你借力 AI、投入自己好奇心的一些方式——我认为最终正是让人朝这个方向前进的脚手架。而且，哪怕只是看听众的评论，我也能感受到它很激励人。就像一条很多人都想走的路，但他们正在努力寻找的是，在自己旅程中该抓住哪些合适的落脚点。

One of the ways that inspires me is just looking at your own role right now. You are a design engineer, and yet – correct me if I’m wrong – there’s not a traditional designer at Paradigm, right? You are kind of the designer, but you’re also owning a lot of the front end. It’s a journey that, in theory, a lot of people listening to this could go on. I actually think that I’m going on it myself. Maybe not as quickly on some of the engineering concepts, but I’m the designer, and I’m just owning more and more of the front end. It can be scary for people who have had a history of code aversion. And yet the blueprint, and some of the ways that you are leaning on the AI and leaning into your own curiosity – I think is ultimately the scaffolding that allows people to progress in this direction. Which, even just from the comments of people listening to the show, I can tell is inspiring. It’s a journey that a lot of people want to go on, but they’re trying to figure out what are the right footholds to grab on their own journey.

在 Paradigm，我非常非常幸运能和一些经验极其丰富、博学的工程师并肩工作。比如我们的前端工程师 Shirley，还有我们的 CTO Jerry。我觉得我一直在向他们学习。作为一个仍然觉得自己在工程道路上相对早期的人，我觉得很大一部分是在摸索：作为在 Paradigm 逐步明确这个设计角色的人，我相对于那些跨层工作的工程师，应该处在什么位置？哪些 PR 由我自己来提交最合适？我觉得核心是要找到正确的抽象层次和边界去平衡。对某些设计师来说，他们真的只想往深处钻研，理解事情背后的原理。而我觉得有些设计师非常擅长把自己脑海中的精确愿景直接表达出来，然后和工程师或任何合作者紧密配合，把那个愿景变成现实。

At Paradigm, I’m very, very lucky to work alongside some extremely experienced and knowledgeable engineers. Like we have Shirley, front end engineer, and Jerry, our CTO. I think I’m always learning from them. And as someone who still feels relatively early in my engineering journey, I think a lot of it has been figuring out, as someone who’s feeling out this design role at Paradigm: what does it look like for me, versus the engineers who are working across all layers of the stack, and what are the PRs that make the most sense for me to make myself? I think it’s really figuring out the right level of abstraction and boundary to straddle. And for some designers, you know, they really just want to tunnel deep into how things work. And I think some designers are incredibly skilled at just spitting out this exact vision that they have, and then working closely with engineers or whoever they’re working with to bring the exact vision to life.

那我们能不能聊聊你的实践？也许我们甚至可以看一个最近的发布，或者甚至只是假设一下：你会怎么处理一件更扎实、更模糊的事情？比如，你会进行多少代码探索？你在多大程度上承担前端实现的全部细节？中间又有哪些步骤？对于一个几乎什么样的工具都能用的人来说，你如何看待自己的工作方式？甚至在过去六个月里，它有没有在某些方面发生演变？

Can we talk a little bit about your practice then? Maybe we could even look at, I don’t know, a recent release, or even just hypothetically: how do you work on something that feels a little bit meatier, more ambiguous? How much are you exploring in code? How much are you owning all of the front end implementation details? And what are the steps in between? For somebody that can reach for really any type of tool, how do you think about the way that you work? And maybe are there certain ways that that has evolved over the last six months, even?

我觉得我们正在摸索的事情是：哪些中间产物是可以精简掉的？哪些事情又值得你花时间，真正做出高保真版本？有时候这甚至不是一条直线。而是要跟团队坐下来，聊一聊到底什么才是正确的思维模型。比如，在 Paradigm，我们最近发布了 workflows 功能。

I think something that we are in the process of figuring out are: what are the intermediate artifacts that you can collapse? And what are the things that you should spend your time really figuring out high-fidelity versions of? Sometimes that doesn’t even look like a straight line. It’s like sitting down with the team and having a conversation about what is the right mental model. So for example, at Paradigm, we recently shipped our workflows feature.

对我们来说，关键在于如何定义工作流的基本单元。我们要把视野拉回到最根本的问题：我们关心的对象是什么？什么定义了工作流？用户在构建什么样的工作流？然后由此延伸，我们需要哪些集成？需要哪些操作？我觉得在我们开始进行视觉实验和原型设计之前，在概念层面上达成一致非常重要，因为我们作为设计师容易掉进的一个陷阱，就是提出一堆方案，然后过早地选定一个。关键是要在范式之内弄清楚该创建的正确事物是什么。一旦我们对具体要做什么达成共识，就可以开始使用 Claude Code 或 Figma 这样的工具。我最近也在试着用 Taper，用来快速摆放一批屏幕，有时保真度各不相同。有时候，你开会只需要用矩形把想法画出来，从概念上讲这就是我想要的对话流程。然后在进入编码原型之前，团队逐步达成更多共识。我们最近尝试过的一件事，是如何展示某一条具体的工作流，然后与工程师合作提交原型 PR，反复测试，最终收敛成今天线上看到的样子。

And for us, it's figuring out how to define the primitives of workflow. It's zooming all the way back out to ask: what are the objects we're concerned with here? What defines a workflow? What workflows are users building out? And from there, what integrations do we want? What actions do we want? I think it's so important to be aligned at a conceptual level before we even start moving into visual experiments and prototypes, because one trap we can fall into as designers is coming up with a bunch of options and then picking one too early. It's about figuring out what the right thing to create is within a paradigm. Once we're aligned on what exactly to make, then we can hop into tools like Claude Code or Figma. And I've recently been playing around with Taper a bit as well, just to put down a bunch of screens, sometimes in varying levels of fidelity. Sometimes all you need for a meeting is to draw things even as rectangles to conceptually show: here's the dialogue that I want. Then progressively we get more alignment as a team before moving into a coded-up prototype. Something we recently played with was ways to show this one specific workflow, then working with our engineers to put up prototype PRs, playing around, testing them, and eventually converging into what we see in production today.

关于工作流这类东西，能不能再具体一点？当你和团队达成一致之后，你最先做的几个视觉产物是什么？为什么做这些？

For something like the workflows, can you go even more specific? Once you were aligned with the team, what were the first few visual artifacts you made, and why?

对于我来说，我仍然喜欢用 Figma 来可视化展示这一交互的各种可能方式。有时我也会把自己正在做的几个版本放进 Variant UI，看看还能探索出哪些别的方向。我非常喜欢 Variant 这款产品的一点是，它像是滚动信息流一样工作。所以一旦我们定义了这次交互中需要看到的约束条件，可能就会出现一个我完全没接触过的模式，然后我会想：哦，有意思，这个可能适合什么场景？因为我觉得关键在于理解用户应该做什么，以及我们能以哪些不同的形式在这种交互中呈现它。所以在我们用一些静态图试过——比如，我希望这个按钮能打开某个溢出对话框的入口——之后，我们会进入原型阶段。这个原型可以直接放在我们仓库里的 Claude Code 中，有时我们团队里也有更偏产品的同事，喜欢用其他工具，比如直接在代码里展示他们设想的交互。我们会截取一些屏幕录像和小片段，互相发 GitHub PR 链接，在暂存环境里测试，感受一下什么更合适，最后再倒推：这正是我们希望在最终版本里呈现的感受。现在我们手里留了一堆用来测试的 PR 链接，就像个小墓地一样。

For me, I still like to use Figma to visualize the different ways we could display this interaction. And sometimes what I like to do is take a few of the versions I'm working with and put them into Variant UI, just to see what other ways I could explore this. What I really love about Variant's product is that it works as a scrolling feed. So once we've defined the constraints of what we need to see in this interaction, there might be a pattern that I completely don't have in my realm of understanding, and then I think: okay, that's interesting. What might be a fitting use case for this? Because I think it's about understanding what the user should be doing, and then what are the different form factors we can serve this up in within the interaction. So after we play around with some static shots—like, okay, I want this button to open up this entry point into an overflow dialog—then we play with a prototype that can either be in Claude Code directly in our repository, or sometimes we have people on the team who are more product-focused and like to play around with other tools, like directly in code, to show the interaction they're thinking about. We'll take little screen recordings and snippets, give each other GitHub PR links to test on staging, and see what feels right before eventually we work backwards and say: this is how we want it to feel in the final version. And now we have this little graveyard of PR links for the tests.

我们再聊聊你在一个非常 AI 原生初创公司里担任设计师的角色。那里有很多独特的挑战。那么请给我们讲讲，产品的特性是如何塑造你的设计实践的？

Let's talk a bit more about your role as a designer at a very AI-native startup. There are a lot of unique challenges there. So tell us a little bit about how the nature of the product has shaped your design practice.

在 Paradigm，我们做的事情是让用户能够一次性向大约 1000 个智能体发出提示。界面用的是电子表格，知识工作者对这个形式非常熟悉，但这完全是另一种思维模型。过去，你定义一条公式，电子表格就帮你把结果填出来。但现在我们把它看作与一群智能体交互的方式。你可以只定义一次工作流中的一组步骤，然后这些列——每一列都有一个提示词，它会在几十、几百甚至 1000 行上执行。所以对我们来说，我们一直在思考的是规模扩张，以及如何把它变成一种在同时运行 200 个智能体而不是只有 20 个时也能感觉很好的交互。我觉得我们很好的一个做法是内部试用，同时思考哪些事情可以并行处理。比如说，我喜欢建列表，收集列表——无论是喜欢的资源还是学到的东西。我会把它们放到一列里，然后创建提示词，看看：这能带出什么相关的新想法？或者你会推荐哪些其他资源？比如，如果我想学动画，那就先建立一个动画资源列表，然后看看还能继续抽出哪些线索。这实际上是在把 AI 工具当作一种帮助自己明确下一步方向的方式。

At Paradigm, we work on a way for users to prompt basically a thousand agents at once. The interface is a spreadsheet, which people are so familiar with as knowledge workers, but it's a completely different model of thinking. In the past, you define a formula and the spreadsheet fills it out for you. But now we see it as a way to interact with a swarm of agents. So you can define a set of steps in your workflow once, and then each of those columns—each column has a prompt that gets executed across tens, hundreds, or thousands of rows. So for us, something we're always thinking about is scale, and how we can bring this to an interaction that feels good when you have 200 agents running versus just 20. I think a really great way we've been dogfooding it is thinking about what things to parallelize. So for me, I love building lists, collecting lists—whether it's resources I love or learnings I have. I'll put them in one column, then create prompts to see: what other ideas can I pull up related to this, or what other resources would you recommend? For example, if I'm trying to learn animation, let's first build up a list of animation resources, and then what threads can I keep pulling? It's really about using the AI tool as a way to help myself figure out where I want to move next.

我想再确认一下规模这部分。你会探索或借助哪些设计可能性来解决这类问题呢？

I want to double check on the scale piece for a second. What are some of the design possibilities you find yourself exploring or reaching for to solve that type of problem?

我们真正聚焦的一点是，如何以非常结构化的方式来表示大规模的事物。当我们回到网格、行和列这种想法时，它非常有道理，因为这是人们已经很熟悉的范式。所以对我们来说，要弄清楚的是：如果我想拥有一个大规模的东西，首先生成这一列项目意味着什么？这需要用一个智能体去搜索网页，帮我构建这个列表。然后从那里开始，我到底需要提取什么？或者说，现在与这个知识体系交互的最优方式是什么？所以我们参考了其他数据密集型的重型产品。过去我们用什么方式表示海量数据？无论是在电子表格里，或者也许在某个阶段看图表、文档等等。我认为关键在于，对于结构化数据，要找出那些不变量。有什么方法能提取出所有数据中恒定不变的东西？无论是你想要的一个字段，还是你定的数据形态，这样 AI 就能真正帮你在搭建过程中把它构建出来。

One thing we've really honed in on is how to represent things at scale in a very structured manner. When we return to this idea of grids, rows, and columns, it makes a lot of sense because this is a paradigm people are already familiar with. So for us, it's figuring out: if I want to have something at scale, what does it mean to first generate that list of items? That's using an agent to search the web and help me build this list. Then from there, what do I even need to extract? Or what is the most optimal way to interact with this body of knowledge now? So for us, we're referencing other very data-heavy products. What are the ways we've represented vast swaths of data in the past, whether in a spreadsheet or, maybe at some point, looking at graphs, documents, and things like that. I think it's about identifying, with structured data, what are the invariants. What are ways you can pull out something that is constant across all of them, whether it's a field you want to have or a shape of the data you can define, so that the AI can really help you build it out as you scaffold it.

你在做非常前沿的技术、工作流和心智模型。你赋能人类工作的方式，跟我们作为一个社会所习惯的实在是太不一样了。我的假设是，作为一家初创公司的设计师，你不得不随时感知周围发生的一切。我们要怎么思考这些变化如何塑造我们自己的产品路线图和可以借鉴的模式？我知道答案肯定是‘刷 Twitter’，但作为设计师，为了健康地跟上世界上发生的一切，你具体在做些什么？希望能从噪音中分辨出信号。

You're working on very cutting-edge tech, workflows, mental models. The way you're empowering humans to work is so different from what we're used to as a society. My assumption is that, given your role as a designer at a startup, you have to keep a pulse on what else is happening around you. How do we think about the ways this shapes our own product roadmap and patterns to pull from? I know the answer is 'scroll Twitter,' but go a little bit deeper into what you're doing as a designer to healthfully stay up to date with everything that's happening in the world, and hopefully be able to identify the signal from the noise.

Twitter 确实会让人非常不知所措，但我认为，保持对大家在讨论什么的感知，就是把它们放进我的待探索清单里。而对我来说，一个真正好的过滤器就是和我的设计师朋友或设计工程师朋友聊天。我有一个想尝试的产品和工具清单，互相交流笔记会让这件事变得更有趣，就像大家一起作为一个集体或社区来做。我非常喜欢为此举办的线下活动，无论是自己主持，还是去参加朋友组织的。它们为人们创造了分享好奇心片段和互相做 demo 的空间：‘这是最近让我特别兴奋的东西，也是我最近一直在深入探索的兔子洞。’我觉得有时候关键就在于挑一件事，用一两个周末深入进去，在那段时间里一边选择深入，一边尝试把事情弄明白。所以，保持广泛涉猎，保留所有这些上下文，然后当某样东西真的抓住你的注意力时，你会想：‘好，现在我必须把它做出来，现在我必须试试看。’然后你就继续回到这样的循环：先铺开，再深入，不断重复。我昨天就和一个朋友这样做了，就是叙叙旧。我们大概有六到十二个月没聊了吧。他是另一位设计师。他在分享他使用 Pencil 的工作流，我在分享我用 Paper 的工作流。看到他在摆弄和试验的东西，真的非常有意思。我们俩都得出了类似的结论：‘天哪，在过去两个月里，我在画布上的探索方式，跟我们职业生涯中任何时候都根本不同。’我们都对此感到非常兴奋。然后他问：‘你试过这个吗？’我说：‘没有。’我又问：‘哦，你试过 Conductor 吗？’他说：‘没有。’能有这样的小型交流真的太宝贵了，因为我觉得我们现在基本上都是在各自开辟道路，边走边摸索。

Twitter can definitely get very overwhelming, but I think keeping a pulse check on what everyone's been talking about is putting it on my list of things to explore. Then, for me, a really great filter is just chatting with my designer friends or design engineer friends. I have a backlog of products and tools I want to try, and comparing notes with each other makes it a much more fun experience to do it together as a collective or community. I really love in-person events for this, whether hosting them myself or going to ones my friends put together. They create places for people to share little snippets of their curiosity and demo to each other: 'Here's something that makes me really excited, that I've been diving down this rabbit hole lately.' I think sometimes it's a matter of picking one thing you want to dive into for a weekend or two and just trying things out for that period while you choose to go deep. So you stay broad, keep all this context, and when something really grabs your attention, you think, 'Okay, now I have to build it out. Now I have to try it out.' Then you just continue to return to that rhythm: go broad, go deep, repeat. I did that yesterday with a friend, just catching up. We hadn't talked in, I don't know, six to twelve months or something. He's another designer. He was sharing his workflow with Pencil, and I was sharing my workflow with Paper. It was so interesting to see the different things he was tinkering with and experimenting with. Both of us had arrived at this similar point: 'Gosh, in the last two months, how I'm exploring on a canvas is fundamentally different than it's ever been at any point in our career.' We were both so excited about it. Then he asked, 'Have you tried this?' I'm like, 'No.' Then I asked, 'Oh, have you tried Conductor?' He's like, 'No.' It's just so valuable to have those little sessions with people, because I feel like we're all just blazing our own trail and figuring it out right now.

绝对，是的。我觉得和朋友分享你的个人工作流真的是一件特别好玩的事，因为你可以深入聊：‘哇，你最近做的这个项目，你是怎么做到的？你在玩什么工具？’而且我一直觉得，保持这种 demo 文化也非常有力量。在加拿大——我想如今已经不止这里了——我们有一个叫 Socratic Coffee 的活动，基本上大家每个周日聚在一起。先一起工作 3 小时，然后最后大家 demo 自己一直在做的东西。这就像一个空间，让你去培养那种做东西的好奇心——不一定要为了工作或某个具体项目，而是专门留出一个空间，对自己说：‘好，今天我要学点新东西，今天我要做点新东西。’仅仅因为热爱过程而去做这件事，我觉得是非常强大的。

Absolutely, yeah. I think getting to share your personal workflows with friends is such a fun thing to do, because you can dive into, 'Wow, this project you made recently, how did you do that? What are the tools you're playing with?' And continuing to have this demo culture, I think it's so powerful as well. In Canada — and I guess by now, well — we have something called Socratic Coffee, where basically people get together every Sunday. They co-work for three hours, and then at the end, people demo what they've been working on. It's about creating spaces where you can foster that curiosity to make things — not necessarily for work or a specific project, but just having dedicated space where you carve out time to say, 'Okay, today I'm going to learn something new. Today I'm going to make something new.' Having that just because you love the process, I think, is such a powerful thing to do.

嗯，这跟你之前谈到的很多事情都有关联。你在刷 Twitter 的时候，有大量内容都是你在说：‘我学到了新东西，这是它。’人们真的很喜欢，真的会产生共鸣。而且现在确实就是这样的时刻。如果你正在收听，并且你受到自己社交媒体旅程和在社区里展示自己的鼓舞，那么现在就是你可以做类似事情的时刻。你只管去探索、摆弄，尝试不同的工具和工作流：‘如果我换一种做法会怎样？如果我把它当作起点会怎样？’然后倾泻出你的收获和体验。人们现在对此如饥似渴。就连我自己的旅程——那也算是我开始的方式——回到 2019 年，Figma 还比较新，它正在变成一个技术性更强的产品，有了很多更强大的功能，比如组件属性。我就会去探索，我会站起身来，用组件做一些奇怪的事情，然后发一条推文。然后我会一次又一次地做。那真的很有趣。我认为同样的机会在当今世界绝对存在，只是因为一切变化太快了。为了乐趣而做事可能是最强大的事情，并且永远保持那种真正的好奇心，不断去建造。比如，工具的形状如何塑造我们这些使用工具的人？它们提供了哪些交互？哪些思维方式真正嵌入到了我们的心理中？你非常喜欢一个工具，就是 Claude Code 的 Ask User 工具，基本上你可以把 AI 当作一个访谈者，来真正帮助你理清自己的想法。所以我认为使用 AI 有两条路可走。一种是你会说：‘好，替我做这个。一次又一次。’然后你陷入一个无限循环，你觉得自己并没有真正开辟新天地，或者你一次又一次遇到同样的 bug。但我真的鼓励我们把它反过来，变成：‘我们如何用 AI 来帮助我们建立正确的心智模型？首先要弄清楚，我应该在哪个抽象层级上工作？为了在这里高效运作，我需要获得哪些必要的上下文？’无论是你真的进入 React，去理解这个组件是怎么工作的，组件应该如何结构化，我应该怎么写这些 prompt，以及它往下传递的方式；还是说，如果我在用 shader 之类的，我真的需要多少数学知识？我认为随着 AI 的发展，我们越来越擅长弄清楚：要做出某个东西，我所需的最小可行知识量是多少。但同时，要成为真正对这项特定技术或这个特定工具驾轻就熟的人，又需要多少知识呢？

Well, that connects back to a lot of the things you were talking about earlier. When you're scrolling your Twitter feed, so much of it is you just saying, 'I learned something new. Here's what it is.' And people really like that, really resonate with that. There's totally this moment in time. If you're listening and you're inspired by your journey on social media and putting yourself out there in the community, this is the moment where you could do something similar. You just explore and tinker and try different tools and workflows: 'What if I did this instead? What if I use this as a starting point?' Then just dump your takeaways and your experience. People are hungry for that right now. Even my own journey — that's kind of how I started, back in 2019 when Figma was a little bit newer and it was becoming a much more technical product with more powerful functionality like component props. I would just explore. I would just stand up and do weird things with components, then make a tweet about it. And I would do it again and again and again. It was fun. I think that same opportunity is definitely present in today's world, just because everything's changing so quickly. Doing things for the fun of it is probably the most powerful thing, and keeping that true curiosity alive and building. Like, how does the shape of our tools shape us as the users of them? What are the interactions they're affording, and the ways of thinking that are really becoming embedded in our psyche? Something you really love is Claude Code's Ask User tool, where basically you can use AI as an interviewer to really help you crystallize your own thoughts. So I think with AI there are two paths you can go down. One is where you're kind of like, 'Okay, just do this for me. Again, again, again.' And you're caught in this infinite loop where you feel like you're not really breaking new ground or you're hitting the same bugs again and again. But I really encourage us to flip it and be like, 'How can we use AI to help us build the right mental models we need? First, figure out what is the level of abstraction that I should exist at. And what is the necessary context that I need to acquire in order to be effective here?' Whether it's, you know, actually going into React and understanding how does this component work, how should components be structured, how should I be writing these prompts and the way it's passing them down. Or, if I'm working with a shader or something, how much math do I really need to know? I think with AI, we're getting much better at figuring out what is the minimum viable level of knowledge I need to make something. But then also, what is the level of knowledge I need to become someone who actually feels well versed in this specific technology or this specific tool.

而且我觉得对每个人来说，这真的取决于具体情境——你只是想做一个有趣的演示或原型，还是这是一个你想长期维持的产品。

And I think for everyone, it really depends on the context — whether you just want to create a fun demo or prototype, or whether this is a product you want to sustain over time.

是的，这又回到了你之前提到的——把 AI 视为终极的个性化导师。

Yeah, it goes back to what you were talking about — viewing AI as the ultimate personalized tutor.

对，没错。

Yeah, exactly.

并且积极寻找方式去拥抱这一点，而不是仅仅完成下一个目标，然后接着下一个目标。

And actively finding ways to lean into that rather than just accomplishing your next goal, and then your next goal.

完全同意。关键是我们如何利用 AI 真正加速我们作为学习者、思考者和设计师的工作方式。

Absolutely. It's about how we can use AI to really supercharge our own process as learners, thinkers, and designers.

我有把握说，你加速了很多人的工作和思考方式——因为你看待这些工具的方式，以及你对待自己旅程的态度和对学习与实验的渴望，都让我们深受启发。所以，我真的非常感谢你今天能来和我们分享。

I feel confident in saying that you have supercharged a lot of people's processes — because the way you think about these tools, but also the way you approach your own journey and your hunger for learning and experimentation, it's super inspiring for us. So I really, really appreciate you coming on and sharing it with us today.

谢谢，Red。这真的是一段非常愉快的经历。

Yeah, thank you so much, Red. It's been such a pleasure.

在你离开之前，我想花一分钟跟你分享一下我最喜欢的产品，因为我总是被人问起我的工具栈。Framer 是我建网站的工具。Genway 是我做研究的方式。Granola 是我在评审时记笔记的工具。Jitter 是我给设计做动画的工具。Lovable 是我用代码实现想法的工具。Mobbin 是我寻找设计灵感的地方。Paper 是我像创意人一样设计的工具。Raycast 是我每一步都会用到的快捷工具。这些公司都是我精挑细选的，正因为有它们，我才能全职做这档节目。所以，支持这个节目最好的方式就是去看看它们。你可以在 dive.club/partners 找到完整的列表。

Before I let you go, I want to take just one minute to run you through my favorite products, because I'm constantly asked what's in my stack. Framer is how I build websites. Genway is how I do research. Granola is how I take notes during crit. Jitter is how I animate my designs. Lovable is how I build my ideas in code. Mobbin is how I find design inspiration. Paper is how I design like a creative. And Raycast is my shortcut every step of the way. Now, I've hand-selected these companies so that I can do these episodes full-time. So, by far, the number one way to support the show is to check them out. You can find the full list at dive.club/partners.