在 Shopify 规模化打磨工艺：Derek Briggs 谈设计作为差异化优势

设计如今成了差异化因素，因为任何人都能想出点子并把它发布出去。但那种品质、用心与热情，是你能从匠人交付的作品里感受到、看到的——他们在意那些额外的细小细节，而只是靠提示词出结果的人则不然。欢迎来到 Dive Club。我是 Rid，这里是设计师永不停歇的学习之地。本周的嘉宾是 Derek Briggs，我们将深入探讨他在 Shopify 担任首席设计师时提升标准的种种方式。他通过屏幕共享展示了他在新版 shop.com 上如何死磕细节，带我们了解他自定义的缓动曲线，以及他在 Shopify 通过设计系统放大影响力的不同方式。那么，让我们直接开始吧。

Design is now the differentiator because anyone can come up with an idea and ship it. But it's that level of quality and care and passion that you can feel and see inside of the things that are shipped by crafters who care about all that extra little detail compared to people who are just prompting results. Welcome to Dive Club. My name is Rid and this is where designers never stop learning. This week's episode is with Derek Briggs and we're going to do a deep dive into all of the ways that he's raising the bar as a principal designer at Shopify. He screen shares how he sweats the details on the new shop.com, walks us through his custom easing curves and different ways that he's scaling impact through designing systems at Shopify. So, let's dive right in.

是啊，上次听到我的消息还是在 Clerk 的时候，和那支团队在一起真是无比愉快。显然那是一群狠角色，我很怀念和他们共事，也一直和其中一些人保持合作，未来还想继续合作。之后我去了 Hypermode 担任设计负责人，和 Kevin Van Gundy 一起。那家公司被收购了，我本来打算休息一阵。但就在那段很短的休息期间，我接到了 Carl Rivera 的联系——当然还有其他公司——但 Shopify 的 Carl 和 Cynthia Sevard，她是副总裁，他是 CEO，想让我加入 Shopify。起初，我记得我跟 Cynthia 说，那不是我这种类型的人适合的公司。我更像创始设计师那一类，我想和小团队快速合作、快速迭代。而她说得很明确：你得跟 Carl 聊聊，我们就是这么工作的，我们正在这样转变，这正是我们想让你加入的原因——因为看到你在其他团队产生的影响力，我们想让你在更大的规模上推动这种影响力，当然，因为我们有很多团队，设计团队大概三百多人，还有几千人的工程团队支持。接到那通电话确实让人发怵，但我对这个机会越来越兴奋。这是一个高主动性、高自主权、能产生影响力的角色，去推动他们的愿景——不只是他们的愿景，还有他们在我身上看到的、与公司愿景的一致：一家极度讲究手艺、极度追求品质、比任何人都更在意细节的公司。这一路很令人兴奋。

Yeah, last heard from me was back at Clerk days and had an absolute blast there with that team. Obviously just a bunch of killers and I miss working with those, have continued to work with some of them and want to continue to work with them in the future. From there then head of design at Hypermode with Kevin Van Gundy. That company got acquired and I was going to take a break. But during that very short break, heard from Carl Rivera among other companies of course, but Carl and Cynthia Sevard at Shopify, she was the VP, he is the CEO, wanting me to come over to Shopify. And initially, I remember telling Cynthia like that's just not the company type that I'm built for. Like I'm more of a founding designer type. I want to work with small teams fast and iterate quickly. And she was very much like you need to talk to Carl, like that's how we're working and how we're shifting things and that's why we want you to be a part of this, because with the impact that we've seen you make at other teams, we want you to come drive that impact at a much larger scale of course, because we have a lot of teams, like the design team's like 300 some people and you know multi-thousand person engineering team to support. So intimidating for sure to get that call, but just got more and more excited about the opportunity. It being a high agency, high autonomy impact role to drive their vision, not only theirs, but the alignment that they recognized in me with their vision of a company that is super high craft, super high quality, cares more about the details than anybody else. It's been exciting.

这中间最大的变化和不同在于，你知道，在创业公司里有很多只是不断尝试，我们在努力找到产品市场契合。我在职业生涯中学到的是——早期我特别执着于细节，也因此打出了名气。但后来，现在变得又老又暴躁了，节奏要快，我们得把那些细节做进去，但也得判断它是否值得。所以有很多快速尝试。但来到 Shopify，这里已经有了产品市场契合，是一家规模巨大、极有价值、极其成功的公司，你开始可以想：我们能做别人做不到的事，因为我们负担得起投入和时间，去达到其他公司负担不起的品质与创新水平。我一直很兴奋于这家公司的另一面会是什么样，那就是我的目标和投入所在。

The biggest change and difference in all that is, you know, in startups there's a lot of just trying things and we're trying to find product market fit. And I've learned that over my career — like earlier on I'm really trying to obsess over the details and it kind of made a name for myself throughout that. But then later on, being old and grumpy now, it's moved quick and we need to get those details in there, but we need to figure out if it's going to be worth it or not. So there's a lot of just trying very quickly. But then coming into Shopify to where there's product market fit, you know, it's a massive, extremely valuable, extremely successful company to where you're able to start thinking about we can do things that no one else can do because we can afford to invest and take the time into reaching levels of quality and innovation that other companies can't afford to do. And I've always been really excited about what does the other side of that look like for the company? And that's where I'm aimed and invested.

我记得几年前和你聊过，之后我还写了一篇小文章，里面有几点我想重新聊聊。第一点是，你谈到在打造这种手艺文化时，自上而下的投入有多重要。你提到了规模，以及在 Shopify 这样的公司工作和在 Clerk 或更创业公司化的环境里工作的不同——在那里你差不多可以直接动手做。我想这里更多是一个“许可”的角度，对手艺和品质的投入必须来自高层。你能谈谈在 Shopify 这样的公司里、在规模化的情况下，那是什么感觉吗？

I remember talking to you a couple years ago and I actually wrote a little bit of an article afterwards and there's a few points that I want to revisit in there. And the first one is you talked about the importance of top-down investment when creating this culture of craft. And you mentioned the scale and how different it is working at a company like Shopify versus something like Clerk or more of a startup environment where you can kind of just do things. I would imagine there's a little bit more of a permission angle here where some of that investment in craft and quality has to kind of come from the top. Can you talk a little bit about what that is like working at scale inside of a company like Shopify?

Carl 还在公司的时候发过一条推文，那是一期叫 Context 的内部播客。有几期已经公开了，那是他和 Toby 之间的一场对话，谈他们如何看待设计在公司下一个时代和下一个方向中扮演的角色。Toby 经常谈到设计就是差异化因素。设计如今就是差异化因素，因为任何人都能想出点子并发布，但那种品质、用心与热情，是你能从匠人交付的作品里感受到、看到的——他们在意那些额外的细小细节，而只是靠提示词出结果的人则不然。即使 Carl 已经不在公司了，Toby 的剧本也没有变。他在我们交付什么、以及由此产生的魔力这件事上，依然非常在意做一个引领者。他经常发布很多这类备忘录，内部备忘录。他很擅长几乎每周都发一封，像是“这是我的一点想法”，一封信。我一直跟团队说的是，这些东西并不一定是在告诉你该做什么。那是他在分享他希望事情变成什么样的愿景，并给你许可去做。而那些最成功的人，就是拿到它就直接跑起来的人——这是一个机会，去说某件事还不够好，然后直接去修好它，直接参与进去。因为自上而下没有人会说：“嘿，嘿，嘿，我们为什么要把它做得更好？”不，没人会这么说。我们有目标，有倡议，有路线图，有需要保持专注的事情，但永远有机会把事情做得更好。品质上的雄心很大程度上是由 Toby 和我们的高层领导驱动的。

Carl put out a tweet back when he was around that was an internal podcast episode called Context. There's been a few of those that have been released publicly, but it's a conversation between him and Toby around how they see design playing a role in the next era of the company and the next direction. Toby talks a lot about how design is the differentiator. Like design is now the differentiator because anyone can come up with an idea and ship it, but it's that level of quality and care and passion that you can feel and see inside of the things that are shipped by crafters who care about all that extra little detail compared to people who are just prompting results. Even though Carl's no longer at the company, it doesn't change the script for Toby. Like he still cares so much about being a leader when it comes to what we ship and the magic that comes from that. There are a lot of these memos, internal memos that he puts out a lot. He's very good at putting out almost a weekly sort of like here's a thought that I have, a letter. What I continue to tell the teams is that these things aren't necessarily telling you what to do. It's him sharing kind of his vision of where he wants things to be and giving you permission to go do it. And the people that are finding the most success are the ones that just take it and run, you know, like this is an opportunity to say something isn't good enough and just go fix it, just go be a part of that. Because no one top down is going to say, hey, hey, hey, like why are we making this better? No, no one's going to say that. We have goals, we have initiatives, we have road maps, we have things that we need to stay focused on, but there's always opportunity to make something better. The quality ambitions are very much driven by Toby and from our senior leadership.

先插播一条消息，然后我们继续。我最近经常直接在手机上做很多构建工作。所以新的 Framer MCP 是一个巨大的解锁。比如说我想对 Dive Club 的 CMS 做一些更新，我只需要给 Claude 或 Grok Bot 发一条语音，就能随时随地做任何修改。这样我既能享受 Framer 在画布上探索想法、敲定响应式设计、更新 CMS 的所有好处，而现在我在 Framer 里能做的一切，我的智能体都能替我做。这是件大事，也是我全力押注 Framer 做 Dive 网站的又一个原因。所以今天就访问 dive.comclub/framer 连接你的智能体吧。好，重大消息。

Real quick message and then we can jump back into it. I've been doing a lot of building right from my phone recently. So the new Framer MCP is a huge unlock. Say I want to make some updates to the Dive Club CMS. All I have to do is drop a voice note to Claude or Grok Bot and I can make any change that I want on the go. So I get all of the benefits of Framer for exploring ideas on the canvas, nailing down responsive design, updating my CMS, but now everything that I could do in Framer, my agents can do for me. It's a big deal and another reason why I'm all in on Framer for the Dive website. So head to dive.comclub/framer to connect your agent today. All right, big news.

现在，你可以在 Claude、Codex、Grok Botter 等任何你工作的地方使用 Jitter，借助全新的 Jitter MCP。我一直说 Jitter 让动画制作比任何其他工具都简单，但这真的把它提升到了另一个层次。比如，你可以在一个关于即将发布的产品会议上，会议结束后立即让 Jitter 开始探索动画社交帖子。或者你甚至可以在一个提示中让设计适应不同的宽高比。而且 MCP 创建的所有内容都是原生 Jitter 文件。所以你不只是被另一个需要从头提示的视频专家困住。你可以轻松介入并手动微调你的动画。这只是 Jitter 成为我动画首选工具的又一个原因。所以如果你想试试新的 MCP，请访问 dive.comclub/jitter。现在进入本期节目。

You can now use Jitter inside Claude, Codex, Grok Botter, wherever you work using the all-new Jitter MCP. I mean, I've always said that Jitter makes animating easier than any other tool out there, but I mean, this takes it to a whole other level. Like, you can be in a meeting about some kind of an upcoming product launch and have Jitter immediately start exploring animated social posts right after it ends. Or you can even adapt your designs to different aspect ratios in a single prompt. And everything the MCP creates is a native Jitter file. So you're not just stuck with another video expert that you have to prompt from scratch. You can easily hop in and fine-tune your animations by hand. It's just another reason why Jitter is my go-to tool for animation. So if you want to give the new MCP a try, head to dive.comclub/jitter. Now on to the episode.

我们能不能聊一聊，或者甚至看一些你可以指出的例子，展示一下你如何利用被赋予的权限来发挥？

Can we talk a little bit about or maybe even look at some things that you could point out as an example for what it looks like for you to kind of run with this permission that's been afforded you?

最近，我进入了 Shop.com 领域。对于不熟悉 Shop.com 的人来说，shop.com 是一个汇集了 Shopify 所有商家的网站。我们有数万家商家，包括极其庞大的、成功的十亿美元级公司商家，也有刚刚起步的人。进入这个领域时，正在进行首页调整。设计已经开始了关于下一个愿景的构思，或者说是下一个迭代。有一些大的变动，然后最终决定我们现在需要的版本与以后需要的版本。因为我们对 Shop.com 有很多非常棒的计划，我非常兴奋能参与其中。我进来时看到这项工作正在进行，我只是觉得还有更多机会，我觉得还有更多机会带来更令人愉悦的体验。这是我最近才跳进去的事情。所以我带来了这个作为例子。但我们在这里看到的是轮播图，一种新的轮播图，已经在 shop.com 上线了。虽然它不错，我可以展示设计，它们略有不同。这是老生常谈的工程实现设计师设计的东西，你希望它尽可能好

So, more recently, I have entered into the Shop.com space. So, for folks that aren't familiar with what Shop.com is, shop.com is a website that is a collection of all the merchants of Shopify. We have, you know, tens of thousands of merchants, extremely extremely large, successful billion-dollar company merchants and also people just getting started. Coming into that, there was a homepage adjustment happening. Design had already kind of started on what the next vision of that was going to look like or guess the next iteration. There was some big swings happening and then ultimately deciding on what's the version that we need now compared to the version that we need later because there's a lot of really incredible stuff that we have planned for Shop.com that I'm extremely excited to be a part of. I'm seeing this work happening as I'm coming in and I just think that there's still more opportunity for I just think there's more opportunity for uh just a more delightful experience. And this is something that I've just recently kind of jumped into. So I had brought this as an example. But this that we're looking at here is the carousel kind of a new carousel that landed on shop.com. While it's nice and I could show the designs, you know, they're slightly different. It's the age-old engineering implemented something that a designer designed and you hope that it's as good

你最喜欢的。

your favorite.

是的。你知道，工程师们很关心，但这需要那种额外的细节水平，我们看到并想要添加进去，你知道我们中的一些人可以做到，当然现在越来越多的人可以使用我们拥有的 AI 工具来做到。但我看到这个，这是我刚接手的第一件事，甚至不是一个倡议。它只是我看到的其他机会之一，产品路线图、人员成长、指导等等。这只是我在侧面进行的事情。但这是现在的版本。它再次不错。它传达了要点，但我只是觉得我们可以为此投入更多体验。设计本身，我和设计师合作，我们想要做，他想要做一个玻璃按钮，我正在想办法如何让这个玻璃在玻璃细节方面更好，这就是实现部分的结果。所以我一直在这里工作，即将发布的是这个版本。所以这个版本我使用 SVG 位移贴图作为背景的滤镜，在下面那个按钮上获得真正的液态玻璃感。但还有更多的事情发生了。你知道，商家的标志，我希望它是最后稳定下来的东西。就像，我们为他们展示商家。我希望它再次成为，那个商家的标志是你最后看到完成的东西，因为它是这个区域最实质的东西是那个名字。这里还有很多其他小细节。我把这个从 Swiper 重写为 Motion 的轮播引擎，这真是太棒了。你知道，Matt 非常关心 motion 包有多好，以及 ticker 组件如此高性能，性能非常重要。但我会指出这里的一些细节。所以，就像我提到的，标志本身是最后稳定下来的东西。我们这里有评分。我不会悬停它，因为悬停会暂停，但评分会微妙地动画到评分值，并与卡片本身一起动画。卡片进来的所有移动都跟踪在它自己的变量上。所以卡片上各个项目的移动基于卡片的移动。我会展示我的意思，当我拖走时，你可以看到东西反向移动。

Yeah. You know, the engineers care a lot and but it just it just requires that sort of extra level of detail that we see and want to add in there and you know some of us can make that happen and more and more people of course these days having access to the AI tools that we have can make it happen as well. But I see this and I'm and this is kind of the first thing that I had just taken on and it's not even an initiative. It's just among other opportunities that I'm seeing product roadmaps, people growing, mentoring, you know, things like that. It's just something that I have kind of going on the side. But this is the version that is now. And it's again it's good. It gets the point across and but I just think there's more of an experience that we can put to this. And the designs itself, I worked with the designer on like we wanted to do he wanted to do a glass button and I was figuring out how can we make this glass like even better as far as glassy like detail and that's kind of where it got ended up with the implementation part of it. So what I'd been working on here that I'm going to get shipped here soon was this version of it. So this version of it I'm using a SVG displacement map as a filter on the background of this to get like a true sort of like liquid glass feel on that button down there. But also, there's just so much more that's happened to this. You know, the logo for the merchant, I want it to be the last thing that settles. Like, this is we're presenting the merchant for them. I want it to be something that again that that logo of that merchant is the last thing that you see finishing because it's it's the most substantial thing in this area is the name of that. There's also so many other little, you know, details in here as well. I had rewritten this from Swiper to Motion's carousel engine, which is just amazing. You know, Matt cares so much about how good the motion package is and with the ticker component just so incredibly performant and performance is so so important. But I'll point out some of the details here. So, like I mentioned, the logo itself just being the kind of the last thing to settle. We have the rating here. I'm not going to hover it because it pauses on hover, but the rating animating up just subtly to what the rating is there and animating with the card itself. All of the movement of the card coming in is all tracked on its own variables. So the movement of the individual items on the card are based on the movement of the card. And I'll show you what I mean by that is when I drag away, you can see that things kind of go in reverse.

那都是跟踪那个单一项目。

That is all tracking that single item.

我可以拉入更多，你可以看到，让我们从这里开始，你可以看到东西如何基于卡片内的那个单一移动变量进来。所以我可以滑走所有这些，你知道，我在这里保持 60 fps 恒定，因为它太高性能了。我可以进入这里，或者点击这里的小项目切换回卡片。这里还有另一件事让我非常兴奋，我有一个 canvas 元素，因为这是我能实现这个的最性能的方式。获取当前图像内渲染的每一视频帧，并在后面播放其模糊版本。所以你可以微妙地看到

And I can pull in more and you can see the uh let's start over here and you can see how things come in based on that single variable of movement within this card. So I can swipe away all these, you know, and I'm holding 60 fps down here constantly just because it's just so incredibly performant. I can go in here or I can click the individual little items here to switch back to a card. There's another thing that's happening here that I'm really excited about is that I have a canvas element because it's the most kind of performant way that I was able to achieve this. taking every video frame that's rendered inside of the image right now and it's playing behind it as a blurred version of it. So you can subtly see

哦，酷。

Oh, cool.

这个视频被模糊处理，就像轮播图的氛围背景光，反映当前上面的内容。如果是图像，就只是渲染图像。但就像这里，你可以看到粉色，你可以看到蓝色。这一切都发生是因为它在后台渲染那一帧，非常高性能，只是为了更沉浸的界面。我经历了很多，比如这些元素中哪个应该先进出。我最终决定在这个交错中，左上角的先走，然后看起来是左下角，然后右上角右下角，但然后反向，你知道这样，以及那种东西。所以这里添加了很多细节，你知道，这是发布的版本，它做到了它应该做的。

this video is blurred as like this like ambient background lighting of the carousel that's reflective of the thing that's on there right now. If it's an image, it's just rendering the image. But like right here in this, you can see the pink, you can see the blues right here. It's all happening because it's rendering that frame in the background as well super performantly to just that more immersive uh interface. I went through a lot of like which one of these elements should should go in and out first. Like I decided up deciding that like in this stagger that the top left one would go first and then it looks like the bottom left and then top right bottom right but then going in reverse you know going this way and how that type of stuff. So there's just sort of a lot of detail added in to where, you know, this is the version that got shipped and it does what it's supposed to do.

它展示的信息其实一模一样，但我们会再扫一遍，琢磨怎么才能真正把这份愉悦感和意图带出来，这也印证了我之前说的、Toby 讲的那句“多上点心”。你就盯着这个东西看，然后你得判断：是什么让它如此不同？是什么让它好这么多？它就是感觉更……更到位。原因就在于那份用心，那份品质。

It displays literally the exact same information, but then we sweep through and figure out how can we really bring the delight and intent out of this, to show again what I was talking about with Toby saying, care more. Just look at this thing and you have to decide, what is making this so much different? What's making it so much better? It just feels so much more. It's just because of that level of care, that quality.

我几乎希望自己能单独看它一眼，好在脑子里反向拆解一下，到底是什么让它感觉这么好。因为你说得对，你越盯着它看，就越会发现里面有那么多细小的决定。比如我一开始根本没注意到那些图片淡出时的错落感，现在你一指出来，我就想：天哪，这差别太大了。还有那个 logo 有多慢。就我自己来说，我不是那种——我不确定自己对缓动曲线有没有那么强的刻意把控，不是我不想，只是它在我脑子里没那么容易开窍。所以有时候我觉得自己就是将就着从模型开箱给出的几个选项里挑一个。我知道我绝对做不出这个 logo 那种延迟收尾的效果，哪怕沾点边都难，但它让人感觉被用心对待了，也让你和现在所有自认为会用 AI 做动画的人区分开来。

I almost wish that I could have looked at it in isolation to try to almost reverse engineer in my mind what is making it feel good. Because you're right, the more that you stare at it, the more there's so many little decisions. Like I didn't notice the staggering of the images fading out at first, and now that you've pointed it out, I'm like, oh my gosh, that makes such a big difference. Or even like how slow the logo is. I think for myself, I'm not somebody who—I don't know if I have as much intentionality with easing curves, not because I don't want to, just because it doesn't click in my mind as easily. So sometimes I feel like I settle for a few different options that maybe the model would give me out of the box and I'm just picking one. And I know that I would not get anything even remotely close to how delayed the resolve is for this logo, but it makes it feel cared for and also separated from everybody who now considers themselves an animator with AI.

我对缓动曲线和这类细节的调校特别挑剔。一条缓动曲线加上它的时长，能做出的效果实在太惊人了。我们都说要把动效和转场控制在 200 毫秒、250 毫秒左右，但你可以用长得多的时长给缓动曲线加细节，同时仍然让它感觉属于那 200、250 毫秒的范畴。我来给你演示一下我的意思。

I'm so picky about easing curves and dialing those types of things. It's just so amazing what you can do with an easing curve and its duration. We all say that we want to keep movement and transitions around 200 milliseconds, 250 milliseconds, but there's amazing things you can do with adding detail to easing curves in a much longer duration, but still making it feel like it's part of that 200, 250 milliseconds. I'm going to show you what I mean by this.

在新的后台改版里，那算是我在 Shopify 接手的第一个项目，大致就是围绕这些东西的视觉语言。我当时特别投入地要在里面建立一套动效系统，让一切都非常一致，和产品里其他所有东西都高度契合品牌。过去一年左右，我在几乎所有 UI 里用的缓动曲线，是这个 cubic-bezier(0.2, 0.8, 0.1)，配 400 毫秒的动效时长。然后你会听到有人说，哦，400 毫秒，那太慢了，那会显得慢。我之前在这里做过一场演讲，展示了我说的意思。你会看到我说的给动效加那些细小的细节、把时长拉长却不影响观感是怎么回事。

So in the new admin redesign, that's kind of the first project that I had taken on with Shopify, was kind of the visual language around this stuff. I was very into creating a movement system within there as well so that everything felt very consistent. It felt very on-brand to everything else within the product. And an easing curve that I have been utilizing in kind of all of my UI over the past year or so is this cubic bezier of 0.2, 0.8, 0.1 with a 400 millisecond movement. And then you hear, oh, 400 milliseconds, like that's slow, that's going to feel slow. And I had done a talk here that showed what I mean by that. And you're going to see what I mean about adding those little bits of detail to movement and lengthening them out but not impacting.

浏览器默认的缓动，你直接套在 transition 上的那个，很机械。它就是移过去了。这是从这里到那里的 200 毫秒，它就这么一下。然后我们还有一个 ease out，同样很机械，但收尾会稍微久一点。所以我默认给所有东西用的 ease out，还是那组 0.2、0.8、0.1 的数值，配 400 毫秒。如果我把这个打开，你就能明白我的意思：它在视觉上仍然按 200 毫秒运作。但如果你盯着它的结尾看，它就有那种——我团队称之为“缓闭抽屉”的效果。就像，你知道的，关抽屉的感觉太舒服了，它滑过去，在最后轻轻缓冲一下。你可以保留这种效果，同时它依然动得很快。视觉上是 200 毫秒，但它留了 200 毫秒给最后那一点点愉悦感，让它变得非常轻柔、自然得多。

So your default browser ease that you just run on a transition, it's robotic. It just moved. This is 200 milliseconds between here and it, and it just does that. And then we have an ease out that again just robotic but it takes a little bit longer to finish here. So my ease out that I use for everything by default is that again 0.2, 0.8, 0.1 is the array of values there, with 400 milliseconds. So if I turn this one on here, you'll see what I mean, that it still operates at visually 200 milliseconds. But if you watch the end of it, it just has this—what my team calls—soft-close drawer. Like, you know, it feels so good to close a drawer and it comes through and it just cushions right at the end. You can have it and it still moves quick. Visually, it's 200 milliseconds, but it reserves 200 milliseconds for that last ending bit of delight that just makes it so gentle and much more natural movement.

如果我们换成 ease out spring，也是一样的道理。这是一个 1,000 毫秒——1 秒长的动画，但因为我把 800 毫秒的移动留给了这个细微的收尾，它就成了一条非常轻柔的 ease out。它在视觉上仍然按浏览器原本那 200 毫秒的动效在走。整体是同时运动的，但结尾有 800 毫秒的视觉细节，让它变得特别悦目。你能看到它是怎么加上缓冲，然后慢慢回滚到位。

If we go to like an ease out spring, it's the same thing. This is 1,000—1 second long animation, but because I'm reserving 800 milliseconds of moving for this subtle little ending to it, it's just this really gentle ease out. It still is visually moving with that original browser 200 millisecond movement. It's all moving at the same time, but there's 800 milliseconds of visual detail at the end of it that just make it so delightful. And you can see how it's adding in the cushion and then that slow roll back to come into place.

说到界面里的动效，我喜欢把它和现实生活中的某个东西联系起来。这有点像你开车停下来的时候，你踩下刹车，慢慢滑行到停止，你能感觉到轮胎往回滚一下。总有那么一个细微的回滚。这本质上就是那个：它进来、收尾，然后只是微妙地落定到位。

And then with the movement in interfaces, I like to associate it to something in real life. And this is sort of like when you're driving a car and you stop, you put on the brakes and you roll slowly to a stop, you feel the tires roll back. There's always that subtle roll back that happens. That's essentially what this is: it's coming in and finishing, and then it's just subtly settling into place.

然后我用不同的策略来决定什么时候用弹簧版本、什么时候不用。通常如果它是撞向某个硬边界，我喜欢让它柔和地进来，然后直接停住。如果它是进入自由空间，我会让它弹一下，让它往上弹一点点。我可以在——比如说，我打开这个内容页——给你演示我的意思。我们 sidekick 输入框这里的这个按钮会触发侧边栏。我给它单独配了一条 ease out，因为我想让 sidekick 落下来的动量，和从侧面进来的东西感觉是连贯一体的。你能看到它在那里发生。当我关掉这个时，它就往外走、向上收尾。所以我在这两者之间切换，感觉它们像是同一个东西的一部分。

And then there's different strategies that I use on when to use like a spring version of something or a non. And it's usually if it's going up against something hard, I like to just have it come in soft and just stop. If it's coming into free air, I'll let it bounce. I'll let it bounce up a little bit. And I can show you what I mean by that on, let's say, I'll go to this content page here. This button here on our sidekick input will trigger the sidebar. And I have a separate ease out that happens because I want the momentum of the sidekick coming down to feel like it's cohesive with the thing coming in from the side. So you can see that happen there. When I close this, then it goes out and finishes upward. So I toggle between those and it feels like they're all kind of part of the same thing.

我在这里不让它弹，因为这里可能有一张表格，可能有某个东西，你不想让它微妙地移动几个像素，因为那可能引发文字换行或截断的位移，进而在页面上造成一些难看的视觉效果。所以这里我是刻意说：我要它柔和地结束，我不要它弹。但页面底部这个悬浮的 sidekick 输入框，它是升到所有东西之上的自由空间里，不会影响到别的东西。所以你会看到我在那里确实用了 ease out spring，只是一个细微的小处理，因为再说一次，它算是升进了那片自由空间。

I don't bounce this here because there could be a table here. There could be something here that you wouldn't want to subtly move a couple pixels because it might cause like a text wrap or a truncation movement, which then visually affects like some ugly stuff on the page. So this here I'm deliberately saying I want it to softly end. I don't want it to bounce. But with this floating sidekick input on the bottom of the page, it's coming up into free air over everything. It doesn't impact it. So you'll see that I do use like an ease out spring on there. Just a subtle little touch, because again it's kind of coming up into that free air.

所以再说一次，这些动画里，你可以在更长的时长里塞进大量细节，同时继续保留那 200 毫秒的视觉要求。再说一次，这是一个 1,000 毫秒、也就是 1 秒长的 ease out spring，但它在视觉上仍然按 200 毫秒的动效在运作。

So again, with these animations, there's so much detail that you can put into a longer duration, but continue to keep that visual requirement of that 200 milliseconds in there. And again, this is a 1,000 millisecond, which is, you know, 1 second long ease out spring, but it still is operating at a visual movement of 200 milliseconds.

我听完这段真的要去改我的规则系统了，因为我就在做同样的事。我把所有东西都卡在 200 毫秒，只因为我想让它感觉干脆利落。但从视觉上看，你说得对。

I'm literally going to update my rule system after this because I'm doing the same thing. Like, I'm capping everything at 200 just because I want it to feel snappy. But optically, you're right.

就像它仍然保持了那种轻快感，只是多了一点额外的精致细节。我的意思是它非常实用，所以谢谢你。

Like it does still keep the snappiness, only it has that little extra detail of polish. And I mean it's ridiculously practical, so thank you.

是的，当然。所以这个：拿不准的时候，就用缓出。超级简单。我不是用 JavaScript 函数来计算这个；这是你直接用 CSS 就能做的。

Yeah, you bet. Yeah, so this one here: when in doubt, ease out. Super simple. I'm not using a JavaScript function to calculate this; it's something you just do with CSS.

再说一次，是 2.801，801 是你的值，配合 400 毫秒，你就到处用它。作为产品 UI 运动的 200 毫秒基线，感觉就是很好。

Again, it's 2.801, 801 are your values with a 400 milliseconds, and you just use it everywhere. It just feels so good as a 200 millisecond baseline for your product UI movement.

我们可以更详细一点，比如缓入缓出。所以缓入缓出：对于像 toast 这样的东西，缓入缓出会是什么样子。更柔和的东西可以出现，但浏览器又像是硬启动和硬停止。在现实生活中，没有什么会那么突然地停止。感觉就是不自然。

We can get a little bit more detailed with like the ease in and out. So ease in and out: this is what an ease in and out would look like for something that might be like a toast. Something more gentle can come into place, but the browser again is just like it's a hard start and a hard stop. And in real life, nothing really stops that abrupt. It just feels unnatural.

所以我们的缓出是 800 毫秒，或者缓入缓出，它专注于那种微妙的运动来开始，然后是一个很好的柔和结束，它实现了同样的视觉运动，就像更微妙的东西，或者不专注于 200 毫秒的目标。我们可以打开像缓出弹簧，它增加了那种微妙的结尾。

So our ease out is 800 milliseconds, or ease in out, and it focuses on that subtle movement to get started and then a nice soft [end], and it's achieving the same sort of visual movement as something a bit more subtle or not focusing on that 200 millisecond target. And we can turn on like an ease out spring, and it's adding that subtle ending.

我喜欢这个结尾。我喜欢这个结尾。

I love the ending. I love the ending.

所以它非常微妙，但就是让人愉悦。

So it's just so subtle, but it just makes it so delightful.

我有一个小小的 Derek Briggs 入门包，我基本上用它来启动每个新项目。那是你关于旧营地按钮以及如何处理阴影和边框之类的推文线程。我现在就要把这个包括进去，你可能只是添加到了我的 Derek 入门套件中，我基本上会永远使用。所以，谢谢你。

I have this little Derek Briggs starter pack that I basically use to kick off every new project. And it's your tweet thread about the old campsite buttons and how you do shadows and borders and stuff like that. And I'm just about to include this now, where I probably just you've added to my Derek starter kit that I'm going to use basically forever now. So thank you.

显然，你已经做了一些分享，因为你提到了人们如何在内部使用这些类型的缓动曲线，但我认为这只是进入另一个我感兴趣的话题的婴儿步，基本上就是：作为一个处于类似首席角色的人，你如何在 Shopify 这样规模的地方产生影响并提升质量？但我的意思是，有数百名设计师，对吧？所以我想象要对你想要对产品产生的那种影响，需要不同的策略。所以你能谈谈这需要什么吗？对于一个感受到 Toby 和团队给予这种许可的人，你如何行动并真正实现你想要的结果？

Obviously, you've done a little bit to share this because you mentioned how people are using these types of easing curves internally, but I think that's like a baby step into another area of conversation that I'm interested in hearing about, which is basically how do you make an impact and level up the quality at a place the scale of Shopify as somebody who's in, you know, you're like a principal role kind of thing, but I mean there's hundreds of designers, right? So it's a different strategy I would imagine to have the level of impact that you want to have on the product. So can you talk a little bit about what that takes, you know, for somebody that feels this permission from Toby and team? How do you act on that and actually achieve the results that you want?

从事管理工作，最近几个角色是设计主管，我喜欢那些。我喜欢建立团队，喜欢在那个层面影响公司。但在 Shopify，我认为与人们脚踏实地地工作非常重要。当我有机会加入公司时，我非常明确地表示我不想管理别人。我想与人们一起工作。

Being in management and again most recently in past couple roles were being head of design roles and I love those. I loved building the teams, loved impacting the company at that level. But in Shopify, I think it's so critically important to be boots on the ground with folks. And I made it very clear when I had the opportunity to join the company that I didn't want to manage people. I wanted to work with people.

对我来说非常重要的一点，不仅从领导力的角度，而且从个人贡献者的角度，是帮助人们了解他们能做什么，并影响他们的职业生涯，学习什么是好的。我也喜欢向人们学习。所以一起建设，会发现一些他们不知道或我没有意识到是显而易见的事情，并帮助他们继续前进，不仅在他们将在 Shopify 内部的其他团队中，而且在整个职业生涯中。

Something that's really really important to me, not only from a leadership perspective, but also an IC perspective, is helping people know what they're capable of and impacting their career, of learning what good looks like. And I love learning from people as well. So building together, there's going to be some sort of thing that's uncovered that they didn't know or I didn't realize was obvious and helping them carry that forward, not only throughout the other teams that they'll be on inside of Shopify, but throughout their career.

加入新团队时对我来说最重要的是，因为我在这家公司里疯狂地跳来跳去。我的意思是，我去过很多团队的角落，与产品团队、销售团队和工程团队合作，只是看看最重要的需要解决的问题在哪里。那就是我要去攻克的。但 upfront 建立信任，我们要做我们最好的工作，这对我来说非常重要，对公司也有影响，这是成功的一大步，因为我看到很多领导者只是要求他们想要完成的事情，或者只是为人们做决定。

The most important thing to me when joining new teams, cuz I jump around this company like crazy. I mean, I've been in so many corners of teams and working with product teams and sales teams and engineering teams and just seeing where the most important problem to solve is. And that's what I'm going to go attack. But building the trust upfront of we're going to do our best work yet is so critically important to me and impactful in the company and it's been a big step of success cuz I see so many leaders sort of just demanding what they want to get done or just sort of making decisions for people.

我最初花很多时间，比如一两天，只是和我将要一起工作的人交流，帮助他们理解我在这里是为了你。我在这里是为了让你一直在做的工作变得更好。暴露更多你兴奋的事情，以及你可能不知道自己有能力做的事情，因为这些会显现并复合到那个产品和那个空间的未来中。

And I spend a lot of time initially talking about like a day or two just getting with the folks that I'm going to be working with and helping them understand I'm here for you. I am here to make the work that you have been doing even better. Exposing more of what you're excited about and what you maybe didn't know you were capable of because those manifest and compound into the future of that product and that space.

我从来不想处于这样的角色：我把事情做得更好，却没有给团队工具、理由、知识，说明为什么它变得更好。我从来不想处于这样的角色：我进来把事情做得更好，然后就离开。比如，影响文化、人们的工作方式、暴露不同的策略和技能组合，也许有人没有意识到他们有能力，这对我来说非常重要，所以当我进入一个项目，花几个月时间，让它变得更好，但我想离开时每个人都为它现在的样子感到自豪，这样如果我进入其他事情，那会继续健康。这非常非常重要。

I don't ever want to be in a role where I'm making something better and not giving the team the tools, reasoning, knowledge of why it got better. I never want to be in a role where I am coming in making something better and just leaving. Like it's really important to me to impact the culture, the way people are working, exposing different strategies and skill sets that maybe somebody didn't realize that they were capable of so that when I come into a project, spend some months on it, make it better, but I want to leave with everybody being proud of where it now is so that if I'm stepping into something else that that continues to just be healthy. That's really really important.

我经常被问到的一个问题是：“你想接管一个团队吗？你想领导一个团队吗？你想围绕这个建立一个团队吗？”诸如此类。我只是认为对于像 Shopify 这样的公司来说，现在这不是最健康的事情。因为 Shopify 越来越自主地成为公司伟大的领域，只是弄清楚需要做什么才能成为迄今为止最好的版本。对我来说，围绕这个影响建立一个团队。当然，我们可以跳来跳去这样做，但我真的想与人们一起工作，我真的想通过向他们展示什么是可能的，而不是谈论什么是可能的，来影响他们的个人贡献。这从来都不是真正让我兴奋的事情。

One of the questions that I get asked often is, "Do you want to take over a team? Do you want to lead a team? Do you want to build a team around this?" Things like that. And I just think that that's not the healthiest thing for a company like Shopify right now. Being that Shopify is becoming more and more autonomously great areas of the company, just figuring out what needs to get done to be the best version of themselves to date. For me to go build a team around this impact. Sure, we can jump around and do that, but I really want to work with the people and I really want to impact the individual contributions of them by showing them what's possible, not talking what's possible. It's just never been something that has really excited me.

我认为从领导力的角度来看，如今需要的领导力越来越少，因为我们拥有所有的 LLM 和 AI 工具，可以立即获得很多答案。你有任何问题，都可以让它发生。我当然抓住了这一点。我运行了很多基于客户如何使用某物的产品机会的自动化，向我展示关于这个没有像你想象的那样转化的想法。所以，这里有一些关于我们如何重新处理这个的想法。就像很多这些循环运行，不断挤出它可能更好的地方。我就是如此痴迷于事情总是不够好。我认为这是我的优点也是我的缺点。

And I think from a leadership perspective that there's just less and less leadership needed these days being that with all of our LLMs and AI tooling that we have, there are so many answers to get instantly. There's any question that you have, you can just make it happen. I've certainly grasped onto that. There's a lot of automations that I run on product opportunities based on how our customers are using something, presenting ideas to me around this isn't converting the way that maybe you think that it would. So, here are some ideas on maybe how we could reapproach this. Like just a lot of those loops running to constantly keep eking out what it could be better. I'm just so obsessed with things not being good enough all the time. I think it's a good thing about me and a bad thing about me.

听你谈到决定更专注于一个几乎更横向的角色，能够触及许多不同的表面领域，这很有趣。很多时候我听到这类角色，它几乎等同于更多地将手艺和你的知识理解系统化，而不是直接与人合作并帮助个人提升。你有多经常思考这种二选一？

It's interesting to hear you talk about the decision to focus more on almost like a more horizontal role almost and being able to touch a bunch of different surface areas. A lot of times I hear that type of role and it almost is synonymous with more systematizing craft and your knowledge and understanding versus working directly with people and helping individuals level up. How much do you even think about that either or?

设计工程目前在 Shopify 内部是一个热门话题，因为还没有专门的設計工程角色。有工程师和设计师，设计师负责看到东西并修复东西。我认为这很棒。但我给你看的轮播图例子，有一定程度的：我明白需要完成什么。我明白我需要提示什么来完成它。而且我明白哪些工具能更好地利用来达到我想要的品质水平，仅仅因为过去构建这些体验、设计和构建这些体验的经验。我仍然认为在这种公司里，这类角色有很大的机会以那种方式施加影响。话虽如此，我与公司内部想要这个的人交谈过，比如领导层想要这个，你知道，嘿 Derek，去那边构建这个东西让它看起来更好。酷。那让那一个东西看起来更好。但如果我们能提取技能、系统和模式，然后部署到整个公司，那么这种心态和技能组合就会无处不在。它不仅仅是在 Derek 所在的地方，或者在其他正在构建的人所在的地方。我认为部署品味、手艺、品质以及关于我们需要使用什么工具和策略来达到我们想要的品质水平的知识，是极其至关重要的。我们需要记录这些，你知道，就用简单的 markdown 文件记录我如何着手、我在做什么，以及这是结果，这样那些技能就可以通过我们的技能包部署到我们的设计师和工程师中，让卓越偶然发生。你知道，就像我希望一切都能偶然变得惊人，因为每个人的智能体都在使用那些真正好的结果曾经用来达到它们所在位置的相同策略，那就是它们被打磨的地方。但然后向前推进，让东西自动变得更好是如此容易。

Design engineering is a hot topic right now inside of Shopify because there isn't sort of a dedicated design engineering role yet. There's engineers and designers and designers are responsible for seeing something and fixing something. And I think that that's awesome. But the example that I showed you with the carousel, there's a certain level of just I understand what needs to get done. I understand what I need to prompt to get it done. And I understand the tools that would be better utilized to achieve the level of quality that I'm looking for just because of the past experience of building these experiences, designing and building these experiences. And I still think that there's a big opportunity for that type of persona in a company like this to influence in that sort of way. With that being said, the conversations that I've had with folks who want this inside the company, like leadership who want this, you know, hey Derek, go over there and build this thing to make it look better. Cool. That makes that one thing look better. But if we can extract skills and systems and patterns that we can then deploy across the company, then that type of mindset and skill set is everywhere. It's not just where Derek is, you know, or where this other person is that you know is building. I think it's incredibly critically important to deploy taste and craft and quality and knowledge around what tools we need to use and what strategies we need to use to achieve the level of quality we're looking for. We need to document those, you know, in just simple markdown files of how I'm approaching this, what I'm doing, and here's the outcome so that those skills can then just be deployed through our skill packages, throughout our designers and our engineers to where greatness accidentally happens. You know, like I want everything to just accidentally be amazing because everyone's agents are using the same strategies that the really good outcomes used to achieve where they were and that's where they were massaged. But then moving forward, it's so easy to make something better automatic.

我喜欢“卓越偶然发生”这个说法，特别是因为有很多设计师来找我，说我觉得自己几乎成了瓶颈或警察，因为其他很多人都能原型设计、能发布，你知道，我在某种程度上被强化质量压得喘不过气来，所以创造成就卓越偶然发生的系统。我真的很喜欢这种心态。

I love the phrase greatness accidentally happens, especially because I have so many designers coming to me asking. I feel like I'm becoming the bottleneck or the cop almost because so many other people can prototype can ship and you know I'm like overwhelmed reinforcing quality in some ways and so creating systems for greatness to accidentally happen. I really really like that as a mindset.

相反，我认为设计师有时需要多花些时间在画板上。我认为有很多提示只是导致智能体为你做什么，而不是你作为人类能做什么。智能体能够做它们学会做的事情，而创新发生在从未发生过的事情中。所以在我看来，用 AI 在设计上创新非常困难，因为它只会呈现过去发生过的事情，而我们试图弄清楚未来会发生什么。所以我认为仍然有机会，LLMs 也可以帮助自动化这些过程。但就像用你的大脑，你知道，用你的技能，用你的品味，用你的雄心坐下来看着某个东西，用手做一些调整，并决定这是我们可以做的，这些是我有的想法。现在实施它。

Contrary I think that designers need to spend some more time on the artboard at times. I think that there's a lot of prompting happening that just results in what an agent will do for you as opposed to what you're capable of doing as a human. Agents are capable of doing what they've learned to do and innovation happens in what has never happened before. So it's very difficult in my personal opinion is it's very difficult to innovate in design with AI being that it's only ever going to present what has happened before when we're trying to figure out what's going to happen in the future. So I think that there is an opportunity still for with LLMs also just helping to automate these processes. But like use your brain, you know, use your skills, use your taste, use your ambition to sit and look at something and make some adjustments with your hands and decide like this is what we can do and here's the ideas that I have. Now implementing that.

是的。我的意思是，这就是设计师比以往更有权力的地方，因为无论这个疯狂的想法是什么，你可以描述它，那个智能体就会在那一刻学习。你知道，你试图实现什么，并帮助你达到那里，它不会停止，直到你满意。但我看到很多人再次在代码中工作，这很棒，但它只和他们能描述的一样好。而不是过去的日子，我们你知道，用像素精心制作高质量界面，能够真正弄清楚那些让界面与众不同的细节，而 LLM 还做不到。它会达到那里吗？是的，可能。但现在在这个时代，我认为设计师更应该真正拥有输出质量，而不是依赖他们的智能体来决定质量。

Yeah. I mean, that's where like designers are more empowered than ever because whatever this crazy idea is, you can describe it and that agent's going to learn at that moment. You know, what you're trying to achieve and help you get there and it's not going to stop until you're satisfied with it. But I see a lot of again folks working in code, which is awesome, but it's only as good as what they can describe it. as opposed to the days when we, you know, crafted high quality interfaces with pixels in place and were able to really figure out those details that set an interface apart that an LLM just can't do yet. Will it get there? Yeah, probably. But right now in this era, I think designers even more should really own what the quality of output is rather than relying on their agents to determine the what the quality.

有一个问题我无法停止问自己。如果公司申请与你交谈，而不是反过来呢？这个问题是全新 Dive Talent Network 的基础。而且它正在起作用。就像现在，我正在帮助我认识的许多最令人兴奋的初创公司招聘收听这个节目的设计师和构建者。所以，如果你好奇外面可能有什么，或者你想加入我的名单，或者你甚至正在寻找你的下一个设计招聘，请访问 dive.club/talent 今天加入。

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. Like 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 and 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.

好的。好的。那么我想稍微跑个题，谈谈你的工作流程在过去一年半到两年里是如何演变的，因为我们聊过，因为你是一个历史上能够拥有大量代码的人，而像我这样的人突然觉得你刚刚获得了这套新技能，而你知道你能够做到这一点，但现在你有了这些智能体超能力之类的。所以，我有点好奇，当你思考如何突破 AI 开箱即用的基线时，你的工作流程是什么样的？

Okay. Okay. So then I want to go down a little quick rabbit trail then and talk a bit about how your workflow has evolved over the last you know year and a half two years since we've talked because you're somebody who has been able to own a heck of a lot of code historically where you know people like me all of a sudden feel like you just got this new set of skills whereas you know you were able to kind of do that but now you have these agentic superpowers or whatever. So, I guess I'm kind of curious like what does your workflow look like when you're thinking about how to push past that baseline that AI is going to give you out of the box?

第一，我总是试图解决客户问题，比如商家和我们的买家遇到什么问题，并研究解决方案是什么？北极星是什么？我们想要达到哪里来完成这个？有太多数据输入可以提取。从分析中提取，从用户使用测试中提取。有太多不同的模式可以用 AI 收集并发现，这里有一些可能转化更好的东西。这里有一些可能转化更好的东西。这里有一些可能转化更好的东西。然后开始精心打造那个结果对我来说是有趣的部分。

Number one, I'm always trying to solve the customer issues, like what are the problems that merchants and our buyers are running into and researching what the solution is there? What's the north star? Where do we want to be to accomplish this? There's just so many data inputs that you can pull from. Pull from analytics, you can pull from user usage testing. There's just so many different sort of patterns that you can collect with AI and uncover that here is something that might convert better. Here's something that might convert better. Here's something that might convert better. And then getting to craft what that result is is the fun part for me.

但就像我说的，我有一些循环在运行，尤其是当某个功能即将上线时，我想知道商家和买家如何使用这个新功能，然后根据我提到的那些数据输入，自动向我呈现机会。所以某个功能上线后，我能看到人们点击、使用它的次数，也许它没有达到我设定的目标。因此，在我专注于其他项目时，它能不断在旁为我创造想法，然后呈现出来，比如“嘿，这是今天关于那个变更的报告，这是它的转化指标”。当然，我们有太多数据可以用来为下一步提供优势，所以我利用大量自动化来为我做这些事——过去我会亲自查看那些百分比和数据点，然后决定下一步怎么做——现在 AI 可以替我做这些，而我可以更专注于我想做的有趣的事情。

But again, like I mentioned, I have these loops that run especially when something is early to get shipped that I want to know how merchants and buyers are utilizing this new thing, and then have it automatically present opportunities to me based on those data inputs that I talked about. So something gets shipped and I can see how many times people are clicking on it, using it, and maybe it's not meeting my goals of what I wanted that thing to be. So to constantly be creating ideas for me off to the side while I'm focused on other initiatives, and then present like, hey, here is today's report on that change and here's the conversion metrics on it. We just, of course, have so much data to use to our advantage on what's next that I'm leveraging a lot of automation around doing that for me—of what I used to do of getting in and seeing what those percentages were and those data points, and then figuring out what to do next—to where AI can just do that for me and I can focus more on the fun stuff of what I want to be doing.

在我们进入有趣的部分之前，我想打断你一下。我其实很喜欢这个答案，因为这是我最近一直在思考的事情：我的实践演进的最大方式之一，是我感觉自己可以访问所有数据。以前我主要依赖别人，因为我不想进去在 Metabase 或我工作的任何地方写一堆 SQL 查询。所以对于听众中那些说“哎呀，我在用 AI 做一堆原型设计，但还没有解锁我实践中的那一部分，而且我没有所有的内部管道”的人——我猜在 Shopify 这应该是相当即插即用的。帮助人们思考那种系统对他们自己来说会是什么样子。比如你收到什么？多久一次？它看起来像什么？我认为这里有一些非常实用的下一步，人们可以自己弄清楚。

I want to stop you before we get into the fun stuff. I actually love that this is your answer because this is something that I've been thinking about a lot recently, where one of the biggest ways that my practice has evolved is I feel like I have access to all the data. Before, I was mostly reliant because I didn't want to get in there and figure out a bunch of SQL queries in Metabase or wherever I'm working. So for somebody who's listening who's like, well shoot, I'm doing a bunch of prototyping with AI but I haven't actually unlocked that part of my practice yet and I don't have all of the internal piping—I'm going to imagine it's pretty turnkey at Shopify. Help people think about what that kind of system could look like for themselves. Like what are you receiving? How often? What does it look like? I think that there's something there in terms of a very practical set of next steps for people to figure out on their own.

外面有太多资源和公司提供数据输入和追踪。他们的整个产品就是为了弄清楚用户在你的产品中的每一次交互。但 LLM 也非常擅长构建你想知道的东西。所以也许你不需要去获得批准实施某个大系统来收集所有这些数据点，也许有些你真正想知道、真正想追踪的东西，你的 LLM 可以为你构建，在一个临时实验中为你获取这些数据。当然，可能涉及到工程部门来决定数据存储在哪里等等，但要 ruthless。我认为人们需要 ruthless 地了解和确信他们设计和交付的解决方案正在实现他们设定的目标。如果你交付的东西没有任何关于其成功的数据，我不理解你怎么能知道你所做的是否对公司正确。所以这就是为什么我对数据以及我的决策对未来的影响非常感兴趣，这样我才能继续迭代，让这件事的成功程度不断提高。

There's so many resources out there and companies that have data input and tracking. That's what their whole product is—to figure out every single interaction that your user is having in your product. But also LLMs are really good at just building the things that you want to know. So maybe instead of having to go get an approval on some big system that you need to implement to gather all of these data points, maybe there's something that you really want to know, really want to track that your LLM can build for you inside of this sort of temporary experiment around getting this data for you. For sure, there might be engineering that needs to be involved with where you're going to store that data and things like that, but just be ruthless. I think people need to be ruthless around knowing and being confident that the solutions that they're designing and shipping are meeting the goals that they set out to achieve. If you're shipping something without any sort of data around its success, I don't understand how you'll be able to know if what you did was the right thing for the company or not. So I think that's why I'm really interested in the data and what impact did my decision make moving forward so that I can continue to iterate on how good the success of this thing can be.

把清晰度再提高一个级别。比如这是通过 Slack 给你的吗？是定时消息吗？还是有问题时你才问？再深入一点，说说现在 AI 某种程度上民主化了数据访问后，你实际与这些数据交互是什么样子。

Crank up the fidelity just one level. Like is this being given to you in Slack? Is it like a scheduled message? Are you just asking questions when they come up? Go a little bit deeper into what it looks like for you to actually interface with this data now that AI's kind of democratized access.

这些循环在我的 harness 中运行，比如在 Claude Code 或 Pi 或 Codex 中。我倾向于直接使用开箱即用的 harness。我坚信我们需要停止用对待昨天 LLM 的方式来对待今天的 LLM，因为它非常擅长做你想让它做的事，而不需要问太多。它更加精准。所以我有的这些定时自动化——我倾向于更多用 Codex 做数据分析之类的，用 Claude 更多做代码输出。这些自动化作为我日常 routine 的一部分运行。坐下来，喝杯咖啡，看看它从昨天的交互中提取了什么，并向我呈现关于关键性指标的想法——比如“这完全不起作用，我们需要尽快做出改变”，或者“我们是不是应该让它再运行一段时间”。大多数情况下是“让它再运行一段时间”。我们不应该每天都对我们该看到什么进行摇摆——你知道，模式会从事物中浮现。人类要么喜欢要么讨厌变化，我们想看看一段时间后会产生什么，但我不想成为监控它的人。我需要把我的精力和热情集中在别处。而这些东西会每天监控它，而且完全不在乎。很多这种兴趣和追踪来自于我直接告诉 Codex、告诉 Claude：我想监控一件事，我希望每天、每周都监控它，这些是我想知道的事情。也许你甚至不知道你想知道什么。你应该问：我应该测量什么？比如这是我做的事情，围绕这个我应该测量什么来证明它是否成功？它会非常擅长给你结果。也许你对它的结果没有信心。那就试一天。我认为通过让它找出你的实现结果有多好，你会学到很多关于你交付的东西。这很多是好奇心——很多是对你能追踪什么、你能完成什么保持好奇。对我们来说，同样，我们有很多不同的 API 连接到我们可以访问的许多不同系统，我可以让我的智能体去拉取信息，并为我准备关于它可能想做什么更改的 PR。这也是另一件事——我绝对不想做它认为需要做的细微修复的工作。我只想审查它。我想这对我来说是下一个阶段：信任其中的自动化部分。但我确实有很多人参与这些事情，我想把解决方案呈现给他们，比如“嘿，如果我们改变这些，你觉得怎么样？”因为我有增长伙伴和产品伙伴，我想让他们参与进来，我想让他们知道我做的更改，因为这也会影响他们的角色。但作为产品设计师，我的角色是呈现这些解决方案，并确保它符合我们正在做的事情的产品愿景，即我们希望它达到的目标。

These loops run in my harness, like in Claude Code or Pi or Codex. I tend to just use the out-of-the-box harnesses. I'm a big believer that we need to stop treating today's LLMs the way that we treated yesterday's LLMs because it's just really good at doing what you want without asking too much of it. It's so much more dialed. So these scheduled automations that I have run inside of—I tend to use Codex more for the data analysis and such, and using Claude more for code output. These automations run as just part of my daily routine. Just sit down with, you know, cup of coffee and go through and see what it pulled from yesterday's interactions and presenting ideas to me on criticality measurements of—this is not working at all, like we need to get a change in here soon, or should we let this bake a little bit. Most often it's let's let it bake. We shouldn't take swings every day on what we should see—you know, patterns emerge from things. Humans either love or hate change, and we want to see what comes from these over a period of time, but I don't want to be the one to monitor it. I need to focus my efforts and my passions elsewhere. And these things will monitor it every day and never care at all. A lot of this interest and a lot of this tracking comes from me just telling Codex, telling Claude, I want to monitor a thing and I want it to be monitored every day, every week, and these are the things that I want to know about it. Maybe you don't even know what you want to know. You should be asking, what should I measure? Like here's the thing that I did. What should I measure around this to prove that this is successful or not? It's going to be extremely good at giving you results around that. Maybe you're not confident in its result. So try it for a day. I think you're going to learn a lot about what you're shipping by just letting it find out how good the result of your implementation was. It's just a lot of curiosity—it's just a lot of being curious with what you can track, what you can get done. For us, again, we have just a lot of different APIs attached to a lot of different systems that we have access to that I can have my agents go pull information from and prepare PRs for me on what changes it might want to make. That's another thing too—I definitely don't want to do the work of the subtle fixes that it thinks need to be done. I just want to review it. I guess that's kind of the next stage for me is trusting the automation parts of that. But I do have many people involved in these things that I want to take and present these solutions to, like, hey, what do you think if we change these things? Because I have growth partners and I have product partners that I want to be involved in this, and I want them to know the changes that I'm making because it impacts their role as well. But as a product designer, it's my role to present these solutions and make sure that it aligns to the product vision of the thing that we're working on, where we want it to be.

是的，这绝对是我设计实践中未曾预料到的部分。很多时候，我只是在查看这些前后对比文档，比如我说“跟踪一个目标”，然后它就会说“好的，开始吧，这是一份报告，顺便说一下，这里有六个提议的想法”，或者我可能会选一个。想想我工作的起点已经向前推进了一些，这真的很有趣，对吧？就像我几乎总是在某些东西的基础上构建，或者至少看着某些东西来做出反应，然后弄清楚“好吧，我想在哪里插入自己？”以及“在这个领域真正提升工艺或交付高质量的东西是什么样子的？”我还在摸索我这部分实践。

Yeah, that's definitely a part of my design practice that I didn't necessarily see coming. A lot of it is just reviewing these before and after documents after I say, like, here track a goal, and then it'll be like, okay here we go, here's a report, by the way, like here are six proposed ideas, or maybe I'm like going to pick one. And it's really interesting to think about the starting point for my work has kind of moved a bit further along in the process, right? Like I'm almost always building on top of something or at least looking at something to kind of react and then figure out, okay, like where do I want to insert myself? And you know what does it look like to actually really level up the craft or ship something that is high quality in this area? I'm still kind of just figuring that part of my practice out.

我电脑上有一个活的智能体配置文件，它基本上就是我的工作方式。我会把与智能体的许多较长对话记录下来，在结束时，我让它把对话总结成主题和我提出的最重要的问题，这些问题影响了我们当时的结果。然后我把这份报告交给另一个附加到我智能体配置文件上的智能体，告诉它将其总结成未来智能体应如何运作的一般主题，这样我就不必再问这些问题了，这样它总是呈现我一直在询问的结果和输出，从而让我越来越高效，不必再问同样的问题，并且它主动获取我明显想要的信息。我做这些事的原因是我想要这些工件，我想要这些 Markdown 文件再次部署到各个团队，这样他们就不必做所有这些工作。我可以为他们完成工作，准备我认为非常有效的解决方案，并把这些工件给他们，这样当 Derek 不在时，这种效率和细节就留给了那些人和那些团队，继续让那个产品领域尽可能好。

I have this living agent profile on my computer and it is essentially how I work. I take a lot of my longer conversations with agents and at the end of it, I ask it to summarize the conversations into the themes and the most important questions that I asked that impacted the outcomes of where we were. And I take that report to then another agent that is attached to my agent profile and I tell it to summarize this into a general theme of how a future agent should operate so that I don't have to ask these questions again, so that it is always presenting the outcomes and the outputs that I am consistently asking about, so that it just makes me more and more efficient and not have to ask the same sort of questions, and it's being proactive on getting me the information that I very clearly want. And the reason why I do these types of things is because I want these artifacts, I want these markdown files again to deploy across the teams so that they don't have to do all this work. I can do the work for them in preparing what I think are very effective solutions to our problems and giving them these artifacts so that when Derek's gone, that type of efficiency and detail is left with those folks and with those teams to continue to make that product area as good as it can be.

你在这个话题上思考得很深入，所以我不想就此打住。如果还有更多要分享的，我想问，回顾过去 6 到 12 个月，在与 AI 合作方面，你觉得有没有其他方式让你解锁了什么，或者在用智能体工作的方式上有所迭代，或者你对这些系统应该是什么样子的正确结构的思考方式有所改变？

You're such a deep thinker on this topic, so I don't want to move on. If there's more to share, I guess is there any other ways when you kind of look back on your last 6 12 months in terms of how you work with AI where you feel like you've unlocked something or iterated on the way that you work with agents or the way that you think about the right structure for what these systems should even be?

我花了太长时间关心代码质量。正如 Toby 所说，我手工雕琢代码很多很多年。我的团队都会告诉你，我非常关心代码组织、代码有多漂亮以及一致性模式。就像，我非常关心我如何编写和雕琢代码。我长期以来都非常在意这一点。直到 Astra 出现。当 Astra 出现，我开始用它写前端时，那是第一次，在没有按照我想要的方式提示它的情况下，我心想：“嘿，这有点像我会写的东西。”那一刻，我意识到我需要停止如此关心这个函数与这个函数命名方式之间的空间以及事物的搭配。

I spent too long caring about the quality of the code. I, as Toby says, hand chiseled code for many, many years. My teams will all tell you I care so much about code organization, how beautiful code is, and the consistency pattern. Like, I just care so much about how I write and chisel out code. And I cared so much about that for so long. And it wasn't until Astra came out. When Astra came out and I started writing front end with it, it was the first time that without prompting it the way that I wanted things to be, that I was like, "Hey, this is kind of like what I would write." And that was the moment where I was like, I need to stop caring so much about this space between this function and the way that this function's named and the collocation of things.

但说到搭配，那正是我开始痴迷的地方。这不太关乎代码如何编写以及它能有多高效，因为我坚信你拥有的最重要的上下文是你的代码库。如果你只是让它继续自我演化，让各种不同的模型做各种不同的事情，把东西扔进去，我认为你最终会陷入巨大的麻烦。所以我想建立这些模式，关于一致性，关于我会写什么，这样它就会写那些东西，但不太关心比如 HTML 可能不像我希望的那么紧凑，它添加了一个我不会添加的额外包装 div，但没关系，会没事的，并放弃那种控制。

But speaking of collocation, that's kind of where I started obsessing. It's less of how code is authored and how efficient it can be because I am a big believer that the most important piece of context that you have is your code base. And if you just let it continue to evolve itself and sling things in there from various different models doing various different things, I think that you're going to get into a world of hurt eventually. So I want to establish these patterns of how that consistency, what I would write, so that it would write those things but care a little bit less around like maybe the HTML isn't quite as tight as what I would like, it added an extra wrapper div that I wouldn't have added, but like that's okay, it's going to be okay, and relinquishing that.

所以，我发现自己在做的是，当我审查代码并看到我希望稍有不同之处时，我告诉我的智能体更新我的智能体 MD 文件，把这些小提示加进去。“就用这个，别用这个。”“我更喜欢你这样做。”“我更喜欢你不要这样做。”诸如此类。但我也在越来越多地梳理和整理这些文件，因为这些 LLM 越来越好，给予更多自由。

So, what I found myself doing is that as I'm reviewing code and seeing something that I wish would be a little bit different, I'm telling my agent to update my agent MD files to add these little nuggets in there. Just use this, don't use this. I would prefer you to do this. I would prefer you not to do this. Things like that. But I've also been kind of combing those and curating those files more and more that these LLMs are getting better and better to give more freedom.

好的。所以我想回到我们第一次对话中谈到的内容，你真的是第一个上节目谈论专职 UI 工程师重要性的人。我想在那之后不久，设计工程师的概念也迎来了它的时刻，我想稍微重温一下这个话题，因为现在你在谈论你的 Astra 时刻。AI 的所有进步如何影响了你如今对团队构成的思考？

Okay. So I want to return to something that we talked about in our first conversation where you were really the first person to come on to the show and talk about the importance of dedicated UI engineers. And I think shortly after that the idea of the design engineer kind of had its moment and I guess I want to revisit that topic a little bit because now you're talking about this Astra moment for you. How has all of the advancements in AI influenced the way that you're thinking about team composition nowadays?

我认为那个角色可能仍然是目前最重要的角色。一个能从代码角度真正将愿景变为现实的人，但这需要那种经验关怀来实现，因为你有这些能写代码的设计师，但如果他们是非技术出身或历史上非技术，他们仍然需要有人在这方面影响他们如何实现他们想做的事情。然后你有这些工程师，他们现在能工作得快得多，把前端组装起来。但你必须要有那个品味制造者，你知道，那个人知道该说什么、如何提示、写什么代码，以及我们试图达到的目标，将两者融合在一起。你有设计的愿景和工程的速度和系统层面。所以，仍然拥有那个负责交付体验的人的支柱，我认为，在如今非常重要。

I think that that role is still maybe the most important role out there right now. Somebody who can truly bring the vision to life from a code perspective, but it takes that sort of experience care to implement that because you have these designers who can write code and but if they're nontechnical or historically non-technical, they still need someone to influence that on how they can achieve what they're trying to do. And then you have these engineers that are working that are able to work much much faster now and get that front end put together. But you have to have that taste maker, you know, that person that knows what to say, how to prompt, what code to write, and the goal that we're trying to reach to blend those two together. You have the design's vision and the engineering speed and system level. So, still having that pillar of that person that's responsible for the deliverable of that experience, I think, is incredibly important these days.

你知道，我们之前在轮播图中展示的，就是一个很好的例子。Astra 可能写了那个。它肯定是最近的。所以我的意思是 5.5 Opus 5.5 或 Astra 构建了那个轮播图，那是我构建的版本，你知道，把设计交给它，你知道，当我进来看到那个机会，我召集团队说“嘿，让我们添加一点细节”，我觉得这可以好得多，但你必须展示什么是可能的。你知道说“它可以更好”是一回事，但展示工件，比如“这是我在想的版本”，你知道设计师和工程师都会，你知道，他们都会说“哇，这感觉好多了”。

And you know, what we shown on the carousel earlier, it's a good example of that. Astra probably wrote that. It was certainly recent. So I mean 5.5 Opus 5.5 or Astra built that carousel, that's the version that I built, you know, get handing it the designs, you know, when I come in and I see that opportunity and I rally the team around, hey let's add a little bit of detail, like I think this could be a whole lot better, but you have to show what's possible. You know saying it could be better is one thing, but to show the artifact of like this is the version that I'm thinking, and you know the designer and the engineer are both going to, you know, they both say whoa this feels so much better.

这里面有更多的意图和细节，但那只是因为我知道我想引入哪些工具，才能达到那里所需的优化和性能基准，还有这种额外的有意义的意图，即如何排序显示。不同的事物根据其价值的权重以不同的速度移动。我们不想在这里放一张图片，而我们可以直接使用 SVG 和滤镜来实现我们想要的效果。有很多这种历史经验，我认为人们能够提供给这些团队，真正帮助他们达到他们想要的标准。

There's so much more intent and detail in this, but that's just because I knew what tools I wanted to pull in that was going to be able to achieve the optimization and performance benchmarks that are required in there, but also this extra sort of meaningful intent of sequencing how it's displayed. Different things move at different speeds based on the weight of its value. We don't want to have an image in place here when we can just use an SVG and a filter on it to achieve what we're looking for. There's just a lot of that sort of historical experience that I think people are going to be able to provide to these teams to really kind of get where they want to be as far as the bar goes.

我认为一个很好的例子是 operate 团队的 Alvish 和 Joe。Alvish 和 Joe 都曾为我工作。他们非常有才华，非常在乎，而 Alvish 并不是技术出身。他为我工作时，一点也不技术。我们只是刚开始涉足那个领域。他——我们一起在 Clerk 和 Hyper Mode 工作过。你知道，他对此充满渴望。他想学习如何做，因为他知道自己有那种品味水平。Joe 也是。你知道，Joe 有那种品味水平，能够实现那些你感受到的体验。所以，现在把他们放在一起。你有 Alvish，他有设计理念，知道东西可以看起来多好，而 Joe 有经验理念，知道东西可以感觉多好，把他们放在一起，看看他们交付的东西。就像，是的，我的意思是，但如果没有 Joe Bell 在 Alvish 的时间线里，谁能达到，你知道，Alvish 知道这个输出有能力达到的，但你必须有一个同样有野心的人，知道我知道它可以有多好，让我们这样做。

I think a good example of this is the operate team with Alvish and Joe over there. Alvish and Joe both worked for me. They're extremely talented, care a ton, and Alvish didn't—he wasn't technical. When he worked for me, it wasn't technical at all. We were just kind of getting into that. He—we worked together at Clerk and Hyper Mode. And you know, he's hungry for it. He wanted to learn how to do it because he knew that he had that level of taste. Joe does. You know, Joe has the level of taste to be able to implement those types of experiences that you feel and whoa. So, you put them together now. You have Alvish that has this design idea of how good something can look and Joe has this experience idea of how good something can feel and you put them together and you look at what they're shipping. It's like, yeah, I mean, but without that Joe Bell in the Alvish timeline, who's going to be able to reach, you know, what Alvish wants knows that this output is capable of, but you have to have somebody that's equally ambitious and knowing that like I know how good it can be and let's do this.

好的，我很喜欢这个例子，因为 Alvish 和 Joe 发布的一切都令人难以置信。就像我的品味库里全是 operate 的截图和 GIF。我今天早上还真的在一封邮件里发了一个。他们在调整可折叠右侧面板的大小，以及列调整大小的方式。我觉得太棒了。不过我要稍微反驳一下，主要是因为我希望你能继续深入探讨一下，在 Shopify 的规模下，这可能会是什么样子，因为 Alvish 和 Joe 的例子太有趣了。当然，如果你把一个非常有才华的视觉设计师和一个非常有才华的设计工程师放在一起，你会得到魔法，他们一直如此。但是当你与数百名设计师合作时，其中许多人我假设可能更多地在写代码，甚至可能为生产做贡献，你知道，世界上没有那么多 Joe，你实际上可以按比例考虑部署 100 个 Joe，对吧？那么，这是否意味着在像 Shopify 这样的组织中，Joe 的价值主张必须更多地倾向于系统，找到创建规则、文档和知识的方法，从而让他们能够扩展影响力，也许目标是让越来越多的 Alvish 能够以 Joe 般的质量直接为生产做贡献，即使它是 80/20。你如何看待这一点？

Okay, I love this example one because everything that Alvish and Joe are putting out is unbelievable. Like my taste library is just full of operate screenshots and gifts. I literally sent one out in an email this morning. They're like resizing the collapsible right panel and the way that the columns resize. I think was amazing. I'm going to push back on it though a little bit mainly just because I want you to kind of keep digging into what you think this could look like at the scale of Shopify because the Alvish and Joe example is so interesting. Of course, if you put a really talented visual designer next to a really talented design engineer, you're going to get magic and they consistently are. But when you're working with hundreds of designers, many of which I would assume are probably working more in code, maybe even contributing to production, you know, there not that many Joe's out in the world where you could actually think in terms of ratios to just deploy a 100 Joe's, right? So, does that then mean that the value prop of Joe inside an org like Shopify has to skew more about systems and finding ways to create the rules and documents and knowledge that then allows them to scale the impact that they could have and maybe the goal is to get more and more of Alvish's being able to contribute directly to production at a Joe like quality even if it is an 8020. How do you kind of think about that?

这完全符合我之前所说的，我们需要把这些工件组合起来，让它们在整个公司范围内扩展，而不是让一个人负责一个领域。如果你有那种技能水平的设计工程师构建这些界面，公司里的其他人想要，他们就可以拥有。这归结为我做了什么？我如何着手？我做了什么让它达到需要的位置？以及我们如何继续保持那种质量水平？然后把这些文件交给那些领域的智能体，你针对某物运行它，它就开始发生。Emil Kowolski 就是一个很好的例子。他一直在发布这些技能，你可以直接像斜杠运行技能，它就会做 Emil 会做的事情。这就是企业级的品味部署，因为你有这些文档，里面有为什么某物如此好的推理，使其如此好所需的工具，然后你知道它需要达到的质量期望才能算好。你不一定需要很多这类人，但你需要那些从系统层面思考如何在整个公司部署这种偏好的人，而不仅仅是在特定领域。我认为公司如果试图雇佣大量这类人来负责所有这些领域，是在浪费时间，而你可以通过 Markdown 文件非常水平地部署他们。

It's exactly aligned to what I was saying earlier around like we need to put together these artifacts that scale across the company rather than having a person responsible for area. If you have that that level of skilled design engineer building these interfaces that other people in the company want, they can have it. It just comes down to what did I do? How did I approach this? What did I do to get it where it needs to be? And how do we continue to maintain that level of quality? and then handing those files off to the agents over in those areas as well to where you run it against something and it just starts happening. Emil Kowolski is a good example of that. He ships these skills all the time around that you can literally just like slash run the skill and it's just going to do the things that a meal would do. And that's enterprise level deployment of taste because you have these documents that have the reasoning on why something is as good as it is, the tools that are needed to make it as good as it is, and then you know the quality expectation of what it needs to be for it to be good. You don't need necessarily a lot of these types of people, but you need ones that are thinking about it in a systematic level of how do we deploy this preference across an entire company, not just in a specific area. I think companies are wasting their time if they're trying to like hire tons of these people to be responsible of all these areas when you can very horizontally deploy them through Markdown file.

这是一个有趣的时期，感觉团队中一些最有影响力的人能够扩展自己和他们带来的东西。我有一个问题，也许我们可以用它作为招聘的过渡。我认为这是你想引入团队的一种人。也许更多是产品设计师，也许更多是关于工艺。我不太确定。但是当你想到你想引入 Shopify 的梦想人选，继续提高工艺和质量的标准时，你具体在寻找哪些东西？比如我们能不能谈谈你对新员工的评估流程，因为我知道你现在正在招聘多个不同的职位。

It's an interesting time where it feels like some of the most impactful people on a team are able to scale themselves and what they bring to the table. I think there's a question that I have that maybe we can even use as a segue into hiring. I think that's one archetype of person that you want to bring onto the team. I think maybe more of the product designer, maybe it is more about the craft. I'm not really sure. But like when you kind of think about the dream people that you want to bring into Shopify to continue to raise the bar for craft and quality and what are some of the very specific things that you're looking for? Like can we talk a little bit about your evaluation process for new hires because I know you're hiring for a bunch of different roles right now.

是的，对我来说非常重要的是，人们不是用 AI 来做 10 倍的工作量。对我来说重要的是，它让他们变得 10 倍更好。所以，他们过去一直手工做的、非常擅长的事情，他们如何能做得更好？比如，他们如何能让输出的质量更好？我也能感觉到有些人变得懒惰了。再说一次，这就是做 10 倍工作量的部分，因为他们就像，“好吧，我可以自动化所有这些。我现在能做的，我可以做一堆。”

Yeah, it's really important to me that folks are not using AI to do 10 times more work. It's important to me that it's making them 10 times better. So, what they used to be doing by hand all of the time and they're extremely good at, how can they make that even better? Like, how can they make the quality of that output even better? I also can sense some folks just getting lazier like that. And again, that's that's the part of the doing 10 times more work because like they're they're like, "All right, I can automate all of this stuff. I what I could do now, I can do a bunch of."

是的，太对了。

Yeah, that's so true.

我们都会这样做，但你把额外的时间用来做什么？我认为对我来说真正重要的是，我把现在腾出来的额外时间用来如何让我的工作更好，而现在就是影响其他人，影响我们团队中其他人的输出，让他们达到他们可能觉得不可能达到的能力。所以，这很多是思想分享，以及我在做什么来激励他们，让他们也开始做更多这样的事情。再说一次，这会在整个公司中复合，带来额外的质量和速度。

And we all do that, but what are you doing with the extra time? And that's I think the thing that's really important to me is that I'm spending the extra time that's now freed up and how can I make my work even better and right now is influencing other people and influencing the output of other folks on our teams to reach their capability that maybe they didn't feel like they were that that was possible. So, it's a lot of sort of thought sharing and and what I'm doing to inspire them so that they can start doing more of this stuff too. And again, that compounds across, you know, additional quality and speed throughout the company.

不过说到底，我在找的是那种特别有渴望的人，他们已经接受了这一点：我们能做成以前从没想过自己能做成的事——不只是靠你以前没有的技能，比如现在你能掌握的技术技能。但“现在你能做”还不够，真正重要的是你意识到自己有能力做出什么。就是那种渴望：“我以前做不了这件事，那我能把它推到多远？”

But yeah, so I'm looking for people that are just hungry, and they've embraced that we are capable of things that we've never thought we were capable of — not only through skill sets that you didn't have before, like technical skill sets that now you can build. But now you can build doesn't feel good enough. It's what you realize you're capable of building that's most important. It's that hunger of "I couldn't do this thing before, so how far can I take it?"

我对这些人特别兴奋。有意思的是，往往是比较年轻的一群人、刚入行的一群人——他们没有被那些老路子困住，也不了解以前是什么样。所以他们做起事来天然就强得多，因为没有旧习惯要打破。而且他们有渴望，有更多时间投入这件事。

And I'm really excited about the people. Funny enough, it tends to be the younger crowd, the newer-to-the-industry crowd — they're not caught in these old ways and they don't understand how it was before. So they're just so much better automatically at what they're doing because they don't have any old habits to break. And they're hungry. They have so much more time to spend on this.

天哪，我都不敢想象 15 年前的我会是什么样——那时我就坐在沙发上逛 HTML Dog，琢磨怎么写 HTML 和 CSS，因为我很沮丧：工程师做出来的东西跟我 Photoshop 文件里的样子不一样。但能活在这个时代真好，简直不可思议。

And gosh, I couldn't imagine where I would be 15 years ago, when I was just sitting on the couch on HTML Dog figuring out how to write HTML and CSS because I was frustrated that the engineers weren't building the things that looked the same as my Photoshop files. But what a time to be alive. It's just crazy.

我觉得这场对话给大家开了很多条小路，让人能把那份渴望用起来，也有不少可以顺着往下追的线索，甚至细到你是具体怎么进阶的。我相信会有几百人拿它当灵感，去琢磨缓动曲线和产品的触感。这里面东西太多了，我真的很感谢你今天来分享。谢谢你抽时间，Derek。

So I think this conversation created a lot of alleyways for people to deploy that hunger and different rabbit trails to go down, even down to just how specifically you leveled up. I'm sure hundreds of people are going to use this as inspiration for easing curves and how products feel. And there's just so much in here that I really appreciate you coming on and sharing today. So thanks for taking the time, Derek.

不客气，兄弟。嘿，如果你喜欢这期节目，别忘了每周四我都会和 Tommy Gio 一起直播 Dive Radio，我们会深入聊热门话题，也会请到业内很多了不起的设计师。想把它加进日历的话，就去 dive.comclub/radio。我们周四美东时间下午 4 点见。

Yeah, my pleasure, man. Hey, if you enjoyed this episode, don't forget that every Thursday I go live with Tommy Gio for Dive Radio, where we deep dive into trending topics and hear from a ton of incredible designers in the industry. So if you want to get that on your calendar, just go to dive.comclub/radio. And I'll see you on Thursdays at 4:00 p.m. Eastern.