在 AI 时代作为通才设计师茁壮成长

与其出去找一堆关于如何构建 Metal 着色器的教程，你直接问 Cursor：“和我一起构建吧。”你在构建，同时也在学习。当你可以直接做原型时，在 Figma 里坐太久或在静态环境里工作太久几乎显得很傻。

Instead of going out and trying to find a bunch of tutorials on how to build metal shaders, you just ask Cursor, 'Build it along with me.' You're building, but you're also learning along the way. It almost feels silly to sit in Figma for too long or work in a static environment for too long when you could just be prototyping it.

你尝到一点着色器的甜头，那就像是世界上最滑的斜坡。我认为总有这种以好奇心为先导的元素，我觉得正是这一点导致了那个时刻。

You get a little taste of shaders and it's like the slipperiest slope in the world. I think there's this element of always leading with curiosity and I think that's what sort of led up to that moment.

通才设计师的价值是不可思议的，对吧？你可以用很多不同的方式做出贡献。

The value of a generalist designer is unreal, right? Like you can contribute in so many different ways.

这是我喜欢 Perplexity 的一点——它非常适合通才，你可以用很多不同的方式灵活发挥。

That's one of the things I love about Perplexity is it lends itself really well to generalists and you can flex so many different ways.

欢迎收听 Dive Club。我是 Rid，这里是设计师永不停止学习的地方。本周的嘉宾是 Gunnar Gray，他负责 Perplexity 的移动端设计。我们将深入探讨设计 AI 产品所需的心态转变，但他也是在 AI 时代作为通才蓬勃发展的绝佳例子。我们聊了很多他制作着色器、设计语音界面和学习编程的经历。但在进入这些话题之前，让我们回到 Gunnar 在 Perplexity 旅程的起点。

Welcome to Dive Club. My name is Rid and this is where designers never stop learning. This week's episode is with Gunnar Gray who's leading mobile design for Perplexity. So we're going to go deep into some of the mindset shifts that are required to design AI products, but he's also a great example of what it looks like to thrive as a generalist in the age of AI. So we talk a lot about his experience making shaders and designing voice interfaces and learning to code. But before we get into all of that, let's go all the way back to the beginning of Gunnar's Perplexity journey.

我刚加入 Perplexity 时，被招来负责移动端设计。那正好是我的专长，我作为唯一的移动设计师加入团队，然后立刻就像被扔进了消防水管。有很多不同的工作轨道，移动应用已经相当不错了，但在完善某些领域和扩展新领域（比如发现 Feed）方面还有很多工作要做。那是我一开始就参与并重新设计的第一批事情之一。

When I first joined at Perplexity, I was brought on to lead mobile design. So that was right in my wheelhouse and I joined the team as like the only mobile designer and then was immediately just like dumped into the fire hose there. And there was a lot of different moving tracks of work and it was already like a pretty good working mobile app, but there was still a lot of work to be done in terms of refining certain areas and then scaling new areas like the discover feed for example. That was one of the first things that I worked on and sort of redesigned right when I started.

你提到了消防水管。能再多谈谈吗？比如最初几个月的生活是什么样的？让我们感受一下在早期 Perplexity 那种规模下做设计师是什么感觉。

You mentioned the fire hose. Can you talk a little bit more about that? Like what was life like in those first few months? Give us a sense of what it's like being a designer at Perplexity in the early days in the midst of all that scale.

是的，一切都在飞速发展，团队仍然非常小。我想我是团队里（包括品牌）第三或第四位设计师。所以每个人当时都在做很多不同的事情，发布速度也非常快。每周都有新东西发布，你加入时就像在追赶一列行驶中的火车。所以，被扔进深水区，然后自己想办法搞定，我觉得这是最好的开始方式。没有太多入职培训，基本上就是拿起东西，跳进 Figma，了解我们在做什么，然后就开始干。所以快速上手、尽快做出贡献很重要，而不是纠结于熟悉过程或学太多东西。边做边学，了解团队如何运作、如何发布、以及产品本身的方方面面。

Yeah, I mean everything was just moving so fast and the team was still very small. I think I was the third or fourth designer on the team including brand. So everyone was doing a lot of different things at that time and the shipping velocity was just going so fast too. So every week felt like something new was shipping and you're just trying to catch the moving train when you hop in there. So, you know, getting thrown into the deep end and figuring it out and making it work. I feel like that was the best way to really get started. There wasn't really much onboarding. It was sort of just like pick up, jump in Figma, you know, understand what we're doing and working on and just go. So, it was important just to pick up quick and start contributing fast and not like dwell on, you know, getting ramped up or learning too much. Like just sort of learning as you go, learning how the team is working, learning how we're shipping, learning more about the ins and outs of the product itself, too.

好了，事情是这样的。你不需要另一个仪表盘。你需要做的是与客户交谈。所以，我想向你介绍 Genway AI。你可以把它想象成一种氛围研究，快速验证你的想法。只需起草你的问题，选择 ICP，然后他们的 AI 智能体就会从全球参与者库中抽取样本，代表你进行访谈。我的意思是，你早上设置好，午饭时间就能得到可操作的见解。验证就在你的指尖，你可以免费试用 14 天。只需前往 dive.club/genway 开始。拼写是 G N W A Y。Raycast 是我的 AI 门户。它只需一个快速的键盘快捷键，这意味着它始终触手可及。我可以添加附件、构建预设以简化某些工作流程，它们甚至支持从 OpenAI 到 Claude 到 Perplexity 的 30 多种模型。所以你的聊天不局限于单一模型。最棒的是你可以免费试用他们的 AI 和所有扩展，无需订阅，无需账户。Raycast 很快就成了我使用电脑的基石。所以，如果你还没试过，帮自己一个忙，前往 dive.com/raycast 开始。拼写是 r a y c a s t。

All right, here's the thing. You don't need another dashboard. What you need to do is to talk to customers. So, I want to introduce you to Genway AI. You can think of it kind of like vibe researching to validate your ideas quickly. Just draft your questions, select an ICP, and then their AI agent runs interviews on your behalf by pulling from a panel of global participants. I mean, you can literally set it up in the morning and get actionable insights by lunchtime. It's validation at your fingertips and you can try it out free for 14 days. Just head to dive.club/genway to get started. That's G N W A Y. Raycast is my portal to AI. It's a quick keyboard shortcut away, which means it's always at my fingertips. I can add attachments, build presets to streamline certain workflows, and they even support over 30 models from OpenAI to Claude to Perplexity. So, your chats aren't bound to a single model. And the best part is you can try their AI and all of their extensions for free. No subscription, no account needed. I mean, Raycast has pretty quickly become the cornerstone of how I use my computer. So, if you haven't tried it yet, do yourself a favor and head to dive.com/raycast to get started. That's r a y c a s t.

好了，现在回到正题。你能更深入地谈谈那些早期的贡献吗？比如，你进入那个环境，节奏很快，几乎没有入职培训，就像“嘿，开始发布吧”。作为设计师，你如何确定自己想要在已经有很多基础功能的产品上产生影响的方式？

Okay, now on to the episode. Can you go a little bit deeper about what those early contributions were? Like, you get in there, there's this pace, it's hardly even an onboarding. It's like, hey, start shipping kind of thing. How did you as a designer then identify the ways that you wanted to make an impact on a product that did already have a lot of functionality that was a baseline that already existed?

是的，它确实有坚实的基础。所以我当时在审计和评估产品，找出我们可以开始改进的地方或发现不一致之处，然后在脑海中开始形成一个设计系统和设计语言，以便更好地定义和推广。这是我关注的一件事。然后为了让事情更容易消化，我做的第一件事之一就是重新设计了设置视图。

Yeah, it certainly had strong foundations. So, I was sort of just auditing and assessing the product and figuring out like where we could start refining things or spotting inconsistencies and just starting to in my own head kind of form a design system and a design language to better define and roll out. So, that was one thing I was looking at. And then to make it a little bit easier to digest, one of the first things I did was just redesign like the settings view.

不错。这是完美的第一步。

Nice. That's the perfect first step.

是的。我的意思是，你了解了产品的一切，所有设置和偏好，然后你可以先在这个隔离区域重新设计，而不触及整个产品。所以这是一个不错的开始，然后我们开始着手发现 Feed，这也非常有帮助，因为它有点像从零到一的项目。我们完全重新设计了它。我们有新的内容创建方式，并希望为参与和消费这些内容创造更沉浸的新体验。这很好，因为我们不必在现有产品或现有模式的太多限制下工作。我们可以为发现创建新的模式和范式。

Yeah. I mean, you understand everything the product has, all the settings and preferences, and you can then just sort of redesign in this isolated area first where you're not touching the full product. So that was sort of a nice start into things and then we just ramped up on the discover feed and that was also really helpful because it was sort of like a new zero to one project. We were totally redesigning the thing. We had new ways of creating content and we wanted to create a more immersive new experience for engaging with that content and consuming it. And that was nice because we didn't have to work too much within the constraints of the current product or any of the current patterns. We got to then create new patterns and new paradigms that we wanted to use for discover.

也许为了给听众一些快速背景，你能谈谈 Artifact，你之前的角色，以及与你设计的东西有多少交叉吗？

Maybe even for quick context for people listening, can you talk a little bit about Artifact, the role you were at before and how much crossover was there with the types of things that you were designing?

是的。在 Artifact，我是作为创始设计师加入的。那时距离上市还有六个月，开始真正为产品定义品牌。

Yeah. So, at Artifact, I was brought on as a founding designer. This was six months away from going to market and starting to really define the brand for the product.

告诉听众创始人是谁，因为这对不熟悉的人来说很酷。

Tell people listening who the founders were just for people unfamiliar with it because it's pretty cool.

是的。Artifact 由 Kevin Systrom 和 Mike Krieger 创立，他们是 Instagram 的创始人。

Yeah. So, Artifact was founded by Kevin Systrom and Mike Krieger, the Instagram founders.

所以能加入那里，和他们一起工作、向他们学习、看看他们是怎么工作的，真的非常酷。我的意思是，他们每天从早到晚都在编码和构建。所以能进来，有机会和他们一起构建，然后帮助他们进一步塑造 Artifact 的设计、打造品牌，并确定我们首先想推向市场的东西，这让我感到无比自由。

So it was really cool to be able to come into there and work with them and learn from them and see how they work. I mean they were in the work all day every day coding and building. So it was so freeing to come in and have that opportunity to just build alongside them and then help them further shape the design of Artifact and work in the brand and sort of identify what we wanted to bring to market first.

而且那里也有 AI 的成分，对吧？我觉得 AI 和 ChatGPT 在我加入 Artifact 几个月后就开始真正起飞了。所以他们做的很多事情是构建所有机器学习和个性化推荐，用于内容发现。然后 AI 开始进入画面，就像现在你可以做所有这些新事情，比如总结内容、改变总结风格、重新格式化甚至重写标题。所以通过重写更好、更清晰、更有价值的标题来避免点击诱饵。这一切都开始迅速加速，我们抓住了这个机会，开始探索我们可以使用它的所有方式，以及我们希望它如何在 Artifact 中呈现。所以和那些早期的想法以及那里的技术一起工作非常有趣。

And there was an AI component to that too, right? I feel like AI and ChatGPT really started taking off like a few months into Artifact when I joined. So a lot of what they were doing was building all the machine learning and personalization for content discovery. And then AI started coming into the picture and it was like now you can do all of these new things like summarizing content, changing the summary styles, reformatting or rewriting headlines actually. So avoiding clickbait by rewriting better headlines that are a little more clear and valuable. So that all started ramping up really quick and we jumped on that and started exploring all the ways we could use it and how we would want that to show up in Artifact. So it was super fun to work with those early day ideas and the technology there.

是的。我的意思是，如果你在 2022 年从事任何与 AI 相关的工作，尤其是在消费领域，那算是 OG 了，这么说有点好笑，但你知道，那确实是很早的时候。所以也许我们可以先高屋建瓴地谈谈，设计 AI 体验是如何塑造你作为设计师的实践方式的。我认为第一件事就是思维模式的转变。我把 AI 原生产品视为一种新的媒介来设计和构建，以及它们与我们过去构建产品的方式有何不同，并思考这种 AI 媒介意味着什么？现在我们有了结果驱动的交互，你只需说出你想要什么，系统就会处理。它非常具有上下文感知能力。所以系统会根据你的请求或要求以完全非确定性的方式动态调整。所以从这个意义上说，它非常流畅。而且它也是多模态的。所以我认为语音在 AI 语境中变得非常重要，因为我们考虑语音界面、对话和语音助手等等。所以我认为你需要理解你所使用的媒介中的一些新基础，以便更好地设计那些输出。所以有那个基础层面，然后还有设计这些结果的过程。以前，从设计角度来看，至少在我的世界里，感觉非常静态，你会花很多时间在 Figma 上，做很多前期工作，然后才开始构建和发布。但自从加入 Artifact，加入 Perplexity，成为一个内部的小型创业团队，环境变得动态得多。现在设计和工程之间的界限非常模糊，有时设计设定愿景，有时工程展示可能性。你需要如此接近那些输入和输出，以至于在这样一个静态环境中工作几乎开始显得愚蠢，因为你希望尽快得到可用的软件，并且希望能够看到那些输出并理解它们是什么，以便按照你的意愿塑造和形成它们。所以你确实需要非常紧密地同时进行设计和构建。我认为这就是 Perplexity 的一点，我们尽可能快地获得可用的软件，这样我们可以学得更快，因为 AI 完全增加了你在一个问题上可以获得的迭代次数，以及你在迭代问题时的速度。所以我觉得那里的反馈循环更加紧密。得出结论的速度快得多，你不想花太多时间闲逛或纠结于想法或设计。你只想做任何必要的事情来让东西工作起来，然后从那里开始改进。

Yeah. I mean if you were working on anything AI related especially in consumer back in 2022 that's kind of OG which is funny to say but you know like that's pretty early days. So maybe we could talk at a high level to start about the different ways that designing AI experiences has shaped the way that you approach your practice as a designer. I think the first thing is just a mindset shift. Like I think of AI native products as a new medium to design and build for and what those differences are from the way we used to build products and thinking about sort of what does this AI medium mean? Now we have outcome-driven interactions where you can just state what you want and the system is going to handle it. It's very much context-aware. So the systems will dynamically adjust based on your requests or asks in a totally non-deterministic way. So it's very very fluid in that sense. And then it's also very multimodal. So I think voice becomes very important in an AI context as we think of voice interfaces and conversations and voice assistants and things like that. So I think there's some newer foundations to understand in the medium that you're working with to better design those outputs. So there's the foundational level of that and then there's also the process for designing those outcomes. So, previously it just felt very very static from a design standpoint in my world at least where you would spend a lot of time in Figma and you would do a lot of work upfront to then start building and shipping. But since joining Artifact, since joining Perplexity, being on an in-house kind of scrappy startup team, it's so much more of a dynamic environment. And there's such a blurry line between design and engineering now where sometimes design is setting the vision or sometimes engineering is showing what's possible. And you need to be so close to those inputs and outputs that it almost starts to feel silly working in such a static environment where you want to get working software as fast as possible and you want to be able to see those outputs and understand what those are in order to shape them and form them however you want. So you really need to be super close like designing and building at the same time. And I think that's one thing at Perplexity where we try to get working software as soon as possible so we can learn much faster because AI has totally increased the number of reps you can get on a problem and as you iterate through a problem. So I feel like there's such a tighter feedback loop there. It's much faster to come to conclusions where you don't want to sit around too long or dwell on ideas or designs. You just want to do whatever you need to do to get something working and then refine it from there.

有没有你发布过的东西的例子？也许是 Discover，也许是别的什么，我们可以深入探讨你们团队如何在这个机会或问题领域内协作，或者你们试图发布的功能。我真的很想了解你在那个过程中如何介入，如何与工程师协作，输出甚至如何随时间演变。我觉得那里有很多可以挖掘的东西。所以我想知道是否有一个例子我们可以指出来。

Is there some kind of an example from something that you've shipped? Maybe it's a discover, maybe it's something else where we could really go deep into how you all collaborate as a team within this bucket of an opportunity or a problem you're trying to solve or a feature you're trying to release. And I really want to understand where you insert yourself in that process, how you collaborate with engineers, how the output even evolves over time. I feel like there's a lot to dig in there. So I'm wondering if there's an example we could point to.

这种工作方式的一个很好的例子可能是我们的语音功能，包括我们的 Android 语音助手，以及 iOS 上的语音模式。那是一个经历了多次迭代的项目，我认为它在测试和实验中反复进行，以真正理解你处理的是什么输出，以及实际的提示是如何工作来获得这些输出的，感觉有多快，性能如何，以及准确度如何。然后还要发现所有不同的用例和所有可能遇到的不同场景，因为语音非常动态。它可能走向任何方向。所以那是这样的：我们设计了核心的壳界面，但它变成了一个更大的系统来定义，涉及来回的语音以及你如何与语音交互，并考虑到你可能处于的所有不同状态。

A good example of that way of working is probably with our voice features, both our Android voice assistant, but also the voice mode on iOS. So that was a project that went through many iterations and I think that it was so back and forth in testing and experimenting to truly understand what outputs you were dealing with and like how the actual prompting was working to get those outputs, how fast it felt and what the performance was like and how accurate it was too. And then also just like uncovering all of the different use cases and all of the different scenarios that you could run into with voice because it's so dynamic. It could go any which way. So that was something that was, you know, we designed like the core shell interface, but it became much more of a system to define in terms of the back and forth voice and how you're interacting with voice and accounting for all of those different states you could be in.

你能多谈谈你作为设计师在塑造语音交互应该变成什么样时所扮演的角色吗？当实际上没有那么多像素参与时。

Can you talk more about the role that you as a designer play when you're shaping what this voice interaction should become when there's really not that many pixels involved?

我认为我在语音模式上的角色更侧重于构建系统，并考虑到你可能处于的所有不同状态，实际上它的界面非常少。但即使界面相当简约，也有大量的反馈和交互需要设计。所以我做的一件事是首先尝试理解所有的交互，以及我们可以映射到哪些模式。所以当你想到语音时，你会遇到几种不同的状态。在它真正激活之前，你会有一个连接状态。你会有一个聆听状态。你会有一个响应状态。当你停止说话时，你会有一个结束状态。那个时刻是什么样的？你怎么知道系统也知道你说完了？所以所有这些小的交互细节实际上是该体验的核心 UI 元素。所以当我进入项目并开始更多地思考它时，我基本上开始定义那些不同的状态，知道它们对体验至关重要。

I think my role on voice mode specifically was more about crafting the system and accounting for all of the different states you might be in where virtually there's very minimal interface to it. But there's so much feedback and interaction to design even though the interface is fairly minimal. So one thing I did was first just try to understand all of the interactions and what sort of patterns we could map to that. So when you think about voice, you're going to have a few different states. You're going to have a connecting state before it's actually active. You're going to have a listening state. You're going to have a response state. You're gonna have an end state when you stop talking. What is that moment like? How do you know that the system knows that you're done talking too? So there's all these little interaction details that are really the core UI elements of that experience. So when I came into the project and started thinking about it more, I kind of went into defining those different states knowing that those are pretty crucial to the experience.

我要深入聊点技术细节，而且是出于私心，因为我正在做这个。昨天我和 AI 来回折腾，试图理清不同的状态和模式，想弄出一个蓝图交给工程师，帮他们优化我们正在为 Inflight 做的底层对话系统。你已经对此有些了解了，所以我就深入一点。当你和工程师合作开发语音产品时，你具体交付什么？在没有太多现成经验的情况下，你如何确定这些交付物，并有效地完善和迭代它们？

I'm going to go really deep and nerdy here and it's going to be for selfish reasons because I'm literally working on this right now. Like yesterday, I was having all kinds of back and forths with AI, trying to figure out different states and patterns and trying to create some kind of a blueprint that I could give to engineers to help refine this underlying conversational system that we're working on for Inflight. So, you've already figured out a little bit about this, so I'm just going to get kind of deep here for a second. When you're collaborating with engineers on a voice-based product, what are the very specific deliverables that you are bringing to the table and what are you then doing to figure out what those should be when there's not as much of a playbook and refine and iterate on those in a way that is helpful.

这正好体现了设计和工程如何紧密合作：设计可以引领愿景，工程也能展示可能性。关键在于尽快做出可用的 demo，然后开始试用，理解它的响应方式，并在真实场景中使用。很容易陷入理想场景和理想用例的陷阱，觉得一切顺利，状态明确，流程清晰。但实际使用中，从真实需求和问题出发，你才会发现那些小边缘案例、细节和真实感受。

I think that's a perfect example of the way that design and engineering work so closely together because design can lead a vision but also engineering can show what's possible. So it's really about getting the working demos as soon as possible and starting to just play with it and understand how it's responding and using it in real context as well. So I think it's pretty easy to fall into the trap of like ideal context and ideal use cases where everything just works and you have the obvious states down and you're like this is exactly how you're going to step through it. But to actually use it in practice and to use it on the daily for yourself like as you want to ask questions or how you want to interact with it. Actually using it from the standpoint of like real use and real problems you have, you start to kind of see those little edge cases or like nooks and crannies and how it's really feeling.

还要理解它的速度感，这对语音来说非常重要。我觉得无论是语音模式还是 Android 上的语音助手，这一点都更深，因为 Android 助手可以做更多事情：连接其他应用、打开网站、执行任务，过程中有很多小细节。所以我们既依赖工程来展示这些时刻和实际运作方式，也要自己测试，逐步发现那些用例、问题或边缘情况。

And also understanding how fast it's feeling too because that's a super important thing for voice. I feel like for both voice mode, but also the voice assistant on Android, that goes even deeper because you can do so many more things on the Android assistant where it can interface with other apps, it can pull up websites, it can do tasks for you, and there's so many little things that can happen during that process. So, we're really both relying on engineering to show us where those moments are and how it's really working, but also really testing it out ourselves and identifying those uses as we go or those problems or those edge cases if that makes sense.

完全理解。我现在明白了尽快做出可试用的代码有多重要，尤其是语音产品，因为设计师不实际使用能做的有限。找出它哪里出问题，然后说：“嘿，这里出错了，我本期望这样，也许我们可以这样设置理想用例。”再反馈给工程师，形成循环。我想你们肯定有很多这样的优化节点。

No, it totally does. And I get now like the importance of racing to get something in code that you can play with, especially for voice is so important because there's only so much you can do as the designer without using this thing. Figuring out where it falls down. Maybe then you're saying, 'Hey, it fell down here. I would have expected this to happen. Maybe this is how we could set up this ideal use case.' And then you're giving that back to engineers and just having that loop. I would imagine you probably had many many of those refinement touch points.

没错。或者当你让它执行任务时，中途发现：“哦，这里其实可以更酷，或者在这个场景下这么做会很有帮助。”关键就是通过实际操作来发现这些细微之处。

Yeah. Exactly. Or if you're asking it to do a task and you realize halfway through it you're like, 'Oh, this would actually be really cool here or this would be really helpful to do in this context.' Like, it's really about going through those motions to uncover all of those little moments.

我有没有跟你提过我那个奇怪的语音模式和 Perplexity 的个人用法？

Have I told you about my weird personal use case for voice mode and perplexity?

呃，没有，应该没提过。

Uh, no, I don't think so.

好吧，我洗碗，对吧？这是我的夜间例行公事。哄宝宝睡觉后，我洗碗，戴上 AirPods，打开语音模式。Perplexity 其实抢走了一些我听播客的时间，因为我随便选个话题，问个问题，然后就会跑偏，最后和语音模式聊上 40 分钟，边洗碗边聊各种疯狂的话题。这其实是我现在主要的学习方式之一。

Okay, so I do the dishes, right? That's like my kind of nightly routine. And I put the baby down. I do the dishes and I throw in AirPods. I open up voice mode and Perplexity has actually stolen a little bit from my podcast consumption because I just pick a topic that I want to riff on and I ask a question and then it turns in this rabbit trail and I'll end up just talking to your voice mode for like 40 minutes while doing dishes and I'll just go down the craziest conversational paths. It's one of the main ways that I learn right now, actually.

太酷了。我也听别人这么用过。这种被动对话很棒，你不需要坐在键盘前，不需要对着电脑或屏幕，可以自由地即兴聊天。

Yeah, that's so cool. I've heard others use it that way as well. It's such a great thing to have that passive conversation where you don't have to be at your keyboard. You don't have to be at a computer. You don't have to be looking at your screen. You can just be riffing like handsfree.

这也一直是语音模式的目标。我们现在基本实现了免提体验，但以前不是这样，你得按按钮。

And that's always been the goal of voice mode as well. And we're pretty much, you know, there now with the hands-free experience. It wasn't always like that. You had to push a button.

那个大橙色按钮。

You had the big orange button.

大橙色按钮。

Big orange button.

好，我们聊聊这个。你刚加入时，语音模式的 V1 已经在移动应用里了吗？

Okay, let's talk about that for a second. Then when you first joined, did V1 of voice mode already exist in the mobile app?

是的，V1 已经存在了。我记得是一个半屏弹窗，覆盖在主视图上。

Yeah, V1 it did exist. I believe it was like a little half sheet that came up. It just kind of overlaid home view. And

所以橙色按钮不是 V1。

So the orange button is not V1.

不是，实际上是 V2。我们有三个版本的语音模式。

No. So actually it's V2. We have three versions of voice mode.

我不知道。好吧，说实话我几乎忘了第一个版本。

I didn't know that. Okay. So I like I honestly kind of forgot about the first version where

我甚至没见过。

I never even saw it. Yeah.

是啊，感觉很久以前了。第一个版本是个半屏弹窗，带一个小正弦波可视化器。那是我加入之前做的，应该也是按讲模式，用的完全是不同的 API，所以我不太清楚第一个迭代的细节。

Yeah. It was like feels like so long ago now. But the first Yeah. It was like a little half sheet. We had this little like sinewave visualizer. It was done before I had joined. I believe that was also like a push to talk. It was using a whole different API and that so I don't know the full ins and outs of like that first iteration.

所以橙色按钮流程是你从 0 到 1 做的，那是你的视觉方向。跟我聊聊那个吧，因为我理解你们做 V2 和免提的所有原因，它很美，但第二个版本从 UI 角度也有独特之处，我很喜欢。我想听听故事，也许我们可以从设计角度聊聊这个功能是如何演进的。

So, you went 0 to 1 on the orange button flow though, like that was your visual direction. Talk to me a little bit about that then, because I understand all of the reasons why you did the V2 and the hands-free and it's beautiful, but there was something also so unique from even just a UI perspective on that second version that I definitely love. So, I would love to talk about it and kind of get some of the story and maybe we can even talk about just at a high level how this feature has evolved from a design standpoint.

第二个版本的语音模式，我们知道它不是最终形态，但必须推出一个能用的、还不错的产品。尽管它没有我们想要的所有交互。所以我们用了按讲按钮作为缓冲，因为语音延迟和问答响应时间有问题。你无法确定模型何时结束，或者它是否知道你讲完了。如果尝试免提模式，它要么过早打断你，要么你讲完后要等很久才检测到。

That second version of voice mode, we knew it wasn't the endgame, but we had to get something out to market, something that worked, something that was still like pretty good. Even though it didn't have all the interactions we wanted. So, we used that push to talk button was, you know, somewhat of a hedge because there were issues with the latency of voice and that ask to response time. You didn't always know like when the model would finish or if it knew you were done talking. If we were trying to do the hands-free mode, it would either interrupt you too early or you would finish and there was like a really long pause before it detected you were done.

所以，通过引入按下通话按钮，实际上让它感觉快了很多，并且带来了一种触觉体验。我们加入了触感和声音。因此，尽管这算是一种折中，我们仍然希望让它感觉非常愉悦，让交互体验良好。它之所以更快，是因为你知道什么时候说完，只需松开按钮就能发出信号。仅此一点就让感觉快了一些，然后整个响应流程就可以立即启动。

So, by introducing the push-to-talk button, it actually made it feel much faster and kind of gave it this tactile experience. We had haptics and sound to it. So, even though it was somewhat of a hedge, we still wanted to make it feel very delightful and we wanted to make it feel good to interact with. So it made it faster because you knew when you were finished and you were able to signal when you were done by just releasing the button. So that alone just made it feel a little bit faster and then the whole response flow could kick in right after that.

我们来聊聊这个按钮，因为没多少人——反正我自己肯定没有——我从未设计过几乎整个界面只有一个按钮的界面。如果只有一个按钮，而且那就是整个 UI，那它必须得是个非常棒的按钮。那么，你探索了多少种方案？你是如何确定那个页面的艺术方向的？

Let's talk about the button for a second because not many people — I mean definitely not myself — like I've never designed an interface where pretty much the entire interface is one button. It's got to be a damn good button if there's one button and that's the entire UI. So, how many things did you explore? How did you arrive at the art direction for that page?

知道页面上有一个大大的按钮，我们得把它做成最好看或最好玩的东西。所以，我们大力投入声音和触感，以及一种与产品其他部分略有不同的新视觉亮点。这样你会觉得这是它自己的特殊事物或特殊实体——这是不同的——而界面的其余部分则建立在这种自适应和动态网格可视化的理念之上。所以我们有一个全屏的点阵网格，我们想用它，因为它能占据整个屏幕，并真正构建更多的视觉反馈机制。我们可以根据你处于什么状态——是正在听你说话、加载中、还是回复你——以不同方式让所有点动起来，然后还有流式文本和图片出现。这基本上是我们想要用于反馈和传达系统状态的基础系统。它让我们能够以酷炫有趣的方式利用图案和色彩来实现这一点，并让语音模式感觉更具沉浸感。无论你是否想看屏幕，它都在那里，供你所需。

Knowing we had a big big button on the page, we got to make this the best looking thing or the most fun to interact with. So really leaning into sound and haptics and just like a new visual accent that looked a little bit different from the rest of the product. So you felt like this is sort of its own special thing or its own special entity — this is different — and the rest of the interface was built on this idea of an adaptive and dynamic grid visualization. So we had this full screen dot grid and we wanted to use that because it was a way to take up the full screen and really build in more visual feedback mechanisms. So we could animate all the dots in different ways depending on the state you were in — whether it was listening to you, when it was loading, when it was responding to you — which then we had streaming text coming in and images. That was sort of the foundational system we wanted to use for feedback and communicating what the system was doing. And it allowed us to do that in cool interesting ways with the patterns and the color and made it feel a little bit more immersive for a voice mode. Whether you wanted to look at the screen or not, it was there if you needed it.

好的。那么现在我们进入 V3。这是一个技术飞跃，你们达到了这样一个点：“好了，免提模式现在感觉可行了。”显然，你得到了一块空白画布，因为它看起来和之前非常不同。那么，你是如何思考适合这种实时语音交互的视觉搭配的呢？

Okay. So then now let's move on to V3. There's the technological jump where you all get to the point where you're like, "Okay, the hands-free mode feels attainable now." You're kind of given this blank slate apparently because it looks very different than what you had. So, how did you think about the right visual accompaniment to whatever this real-time voice interaction could become?

是的。实时性确实改变了它。它变得更快、更直接。你几乎可以瞬间从提问到得到答案。而之前那个全屏大可视化器，在它快速变化时可能会感觉有点突兀。所以我们看了几个不同的版本，并借助 Cursor 这样的工具快速制作了一些 Metal 着色器，看看在那个环境中我们能做什么，以及能做出哪些有趣的可视化效果来帮助传达系统反馈或不同状态。我们用网格做了一些探索，但同样，它占用的表面积太大了。然后我最终确定了这个带有 Metal 着色器的粒子球体可视化器。它立刻感觉像是语音的完美形态，因为它可以根据你所在的状态以多种方式变形。它是屏幕上非常聚焦的实体。它可以用截然不同的方式变形，比如一个完整的圆形球体，或者更像一个环形可视化。所以不同状态之间的那种对比感觉恰到好处。然后加入越来越多的颜色——比如提问或聆听时用不同颜色，回复时用另一种颜色——它移动和变化的方式感觉非常动态和自然，而且比我们以前的全屏网格体验更柔和一些。所以我认为它非常适合语音和实时功能，而且在我们进入更多状态之间的转换时效果很好。我们引入了更多的助手和任务。所以你可以让语音模式为你调出可操作的小部件——预订餐厅、播放播客、播放 YouTube、查找地点，各种不同的用例。你可以看到这个粒子球在不同状态之间转换，它会向下移动一点，稍微弱化，以便其他信息可以接管并拥有一个干净的表面。所以这感觉确实是一个正确的模式，然后就是真正微调 Metal 着色器，并与生产团队合作将其集成并实际连接到语音。这是我第一次真正跳入代码和原型设计，制作出工程团队可以接手并投入生产的有用东西。

Yeah. Real time really changed it. So, it changed it in that it was so much faster and it was so much more direct. So, you were able to go from ask to answer almost instantly. And the previous version of this big full screen visualizer, it was going to feel a little bit jarring maybe in how fast it could sort of move. So, we looked at a few different versions and jumped into tools like Cursor to spin up some metal shaders to see what we could do within that environment and what sort of interesting visualizers we could make to help communicate that system feedback or those different states. And we did some exploration with the grid, but again it was almost like taking up too much surface area for what we wanted. And then I ended up landing on this particle sphere visualizer with metal shaders. And that instantly felt like the perfect form factor for voice because it could sort of shapeshift in many different ways depending on the state you're in. And it was a very focused entity on the screen. It could shape in different ways that were drastically different too. It could be like a full circle sphere or it could be more of a ring visualization. So having that contrast between the different states just felt right. And then layering in more and more color — like different colors for asking or listening and a different color for the response — just the way it could sort of move and shift around felt really dynamic and natural and a little bit softer than the full screen grid experience that we used to have. So I think that lended itself really well to voice and the real time function and it really worked well when we got into transitioning between more states. So, we introduced more assistants and tasks. So, you could have voice mode bring up actionable widgets for you — booking restaurants, playing podcasts, playing YouTube, looking up places, all sorts of different use cases. And you could really see this particle orb sort of transition between the states where it kind of moves down lower and it's a little deemphasized so that other information could take over and have a clean surface to do so. So it really felt like the right pattern and then it was about really dialing it in on the metal shaders and working with production to get that in and actually hooked up to voice. So it was my first sort of intro into actually jumping into code and prototyping and making something useful that engineering could roll with and roll it into production.

你当时是在用着色器做探索吗？

Were you doing explorations with shaders?

嗯，是的。

Yeah. Mhm.

跟我聊聊那个过程吧。

Talk to me a little bit about that process.

感觉非常赋能——我觉得像 Cursor 这样的工具完全降低了入门门槛，对于任何没有编程经验或想学习编程或原型设计的人来说。现在门槛非常低，能够打开 Cursor，不仅提示它做事，还能让它向你解释发生了什么，这感觉非常赋能。所以把它当作学习机制——而不是出去找一堆关于如何构建 Metal 着色器的教程，你只需让 Cursor 和你一起构建。而且你还可以问问题，比如解释某个东西是如何工作的，或者给你分解这个和那个的作用。所以你在构建的同时也在学习。我觉得这对我来说是一个巨大的突破。而且我一直在越来越多地使用它，无论是在任何原型设计场景中，还是试图在 Perplexity 更多地投入生产。

It felt so empowering to be able to — I feel like tools like Cursor have just totally lowered the barrier to entry for anyone who hasn't coded or wants to learn how to code or learn prototyping more. That barrier is so low now and it felt really empowering to be able to just open up Cursor and not only prompt it to do things but also have it explain to you what was happening. So using it as a learning mechanism — instead of going out and trying to find a bunch of tutorials on how to build metal shaders, you just ask Cursor to build it along with me. But also you could ask questions about explaining how something is working or giving you a breakdown on what this does and what that does. And so you're building but you're also learning along the way. And I felt like that was a really big unlock for me. And it's something I've continued to use just more and more in any prototyping context or even like trying to get more into production at Perplexity.

你显然也上瘾了，因为我刚看到你现在有一个用着色器做的小副业项目。

You obviously got hooked, too, because I was just looking and you have a little side project now working with shaders.

是的，我上瘾了，老兄。我现在总是尝试用代码构建。我尽量摆脱 Figma，因为我尝到了直接用代码构建的滋味，并且真正看到了其中的优势。所以，当你可以直接进行原型设计时，长时间坐在 Figma 里或在静态环境中工作几乎感觉有点傻。

Yeah, I'm hooked, man. I'm always trying to build in code now. And I'm trying to get out of Figma as much as I can because I got a little taste of what it's like to just build in code and I've really seen the advantages there. So it almost feels silly to sit in Figma for too long or work in a static environment for too long when you could just be prototyping it.

在你看来，有哪些优势？我觉得我们从宏观层面听到了很多，而且我看到很多人会很快地说，当然领导们希望我们更多地写代码，因为他们只是想少付钱。但作为一名独立贡献者，你看到的优点是什么？

What are some of those advantages in your mind? I think we hear a lot about it from a macro perspective and I see a lot of people that are quick to say, well, of course leaders want us to get more into code because they just want to pay less people. But as an IC, what are the pros that you see?

优点就是速度。它绝对更快，更容易使用，学起来也更快。所以，能够一边编码一边感受交互，会让你做出更好的选择，或者很快理解某个东西是否行不通。同时，你也在系统的边界内工作，了解哪里可以推进、哪里可以调整、约束在哪里、限制在哪里。这样你就不会花太多时间纠结于一个解决方案，而这个方案可能一上线就得改。所以，这确实加速了整个过程和迭代周期。而且，最终你有一个能用的东西可以交给工程团队，他们可以直接接手，这省去了他们大量时间去理解你静态设计稿甚至原型的意思，因为最终你有了可运行的代码，可以直接投入生产。他们当然可以重写一部分，但比起一个原型视频或带注释的静态设计稿，交出一份可运行的软件要有用得多。

Pros are just speed. It's definitely faster. It's easier to work with. It's faster to learn, too. So really being able to code and feel out the interactions at the same time just leads you to make better choices or understand very quickly if something's not working. And you're also working within the bounds of the system as well and understanding where you can push and where you can pull and where your constraints lie or where those limits are. So you're not spending too much time dwelling on one solution that you might end up changing anyways as soon as you get to production. So it's really just speeding up that process and that iteration cycle. Also, at the end of the day, you have something working that you can hand off to and engineering can roll with that and it saves them a ton of time trying to figure out what you mean by like a static comp or even a prototype because you have working code at the end of the day and you can then just roll that into production. They can rewrite some of it, of course, but it's still it's much more helpful to have like a piece of working software to hand off versus maybe like a video of a prototype or some static comps with notes.

嗯，所以这也确实帮助了设计与工程之间的协作。

Um, so it's really helped the that engineering collaboration on design as well.

关于着色器，我做过很多 3D 和 After Effects 的工作。这是过去四年里我出于好奇而自学的东西，因为我真的对它的工作原理感兴趣，并且被它能创造的一切以及带来的机会所启发。我确实考虑过用 3D 和 After Effects 来制作着色器体验的原型，但最终我还是觉得这没法真正交付。而且它可能看起来也不完全是我想要的样子。我宁愿在系统环境中理解它的工作原理，然后从那里开始进行艺术指导。所以所有的艺术指导都是通过即兴创作和用 Cursor 调整不同的着色器模式，然后调优、打磨成我脑海中的样子。当然，我本可以在 After Effects 和 Cinema 4D 里做，那对我来说会更自然。但我想要一个更有用的成品，而且我也想学习。我只是好奇它在 iOS 的 Metal 着色器环境中如何工作。现在我们有能力做到这一点。学习一个新工具并在那里使用它，这是很合理的。但每次我接触新东西时，关键就是哪个工具能最快地完成工作。所以，有时是静态设计稿，有时是某个孤立的组件，有时是动画视频——如果我真的没法用 Cursor 或其他原型工具实现的话。所以，我真的把这些工具都看作工具箱里的工具，根据你要做的事情来选择。但你应该思考什么是最快的路径，以及如何最好地使用这些工具。

I think with the shaders, I've done a lot of 3D work and after effects. It's something I've just picked up out of curiosity over the last four years because I was genuinely interested in how it works and I was inspired by everything you could create and sort of what opportunities that opens. So I did think about like 3D and After Effects as a way to prototype the shader experience, but at the end of the day I was still thinking like I can't really hand this off. And it also might not look exactly how I want it to. Like I would better understand how it works in the system in the environment to start art directing from there. So all the art direction came in just riffing and cursor on different shader patterns and sort of tuning those and refining those to what was in my head where certainly could have done it in After Effects and and Cinema 4D and that that would have felt more natural to me. But I wanted it to feel like a more useful artifact and I wanted to simply just learn too. Like I was just curious how it could work in a metal shader environment in iOS. Now we have the the ability to do that. It just made sense to to pick up a new tool and and learn it there. But every time I'm, you know, approaching something new, it's really just about like what tool is going to get the job done the fastest. So, like sometimes it's a static mockup, sometimes it's a just one isolated component of something, other times it's an animated video if if I really can't like do that or pull it off in cursor or with um a different prototype. So, really, I see all these tools as just exactly that, just tools in your toolkit, and you get to draw from those depending on what you're trying to do. But you should think about like what will be the fastest path forward and and how to best use those tools.

我知道你有视频背景。在 Perplexity 之前，你的技术背景有多深？

I know you have the video background. How much of a technical background did you have before Perplexity?

技术背景并不多。早期我做一点网页设计时，做过一些开发，比如 HTML、CSS、一点点 JavaScript，但那都是很多年前的事了。我已经很久没碰那些东西了。所以，我感觉自己基本上是从零开始，尤其是进入 Swift UI 之后。那是一种不同的语言，在 Xcode 里也是不同的工具。所以，那一切都是新的，但 Cursor 在过程中给了我很多指导，帮了我很多。

Not a ton of technical background at all. So early on when I was doing a little bit of web design, I did a little development like HTML, CSS, little bit of JavaScript, but that's years behind me. I haven't picked that stuff up in a long time. So, I felt very much like I was starting from scratch and especially getting into Swift UI. That's a different, you know, a different language and a different tool in Xcode. So, yeah, that was all new, but Cursor was really helpful to sort of guide me along the way and and help me out.

我真的很希望你会这样回答。是的，因为我觉得做这个播客我最喜欢的部分就是——我记得某个时刻，感觉既像两周前又像五年前，所以我完全不知道具体是什么时候，但所有 AI 工作流和工具都涌现出来，我自己——我相信很多其他人也一样——感到一点 FOMO，一点畏惧，比如“天哪，我要被落下了。这一切太可怕了，我不想跳进去。”然后一个接一个，我听到像你这样的人来分享说，“是啊，我完全不知道怎么做这个。”然后我出于好奇，就试了一下。然后，你知道，现在你把这个惊人的着色器交互交付给了科技界最流行的移动应用之一。这太鼓舞人心了，尤其是考虑到你的背景，天哪，哪怕只是过去六个月，你就能做出这些事情。

I was really hoping that was going to be your answer. Yeah, because I think that's like my favorite part of being able to do this podcast is I remember a moment in time simultaneously feels like two weeks and 5 years ago, so I have no idea when it actually was, but all the AI workflows and tools were coming up and myself, like I'm sure many other people felt this little bit of FOMO, a little bit of intimidation like, "Oh my gosh, I'm getting left behind. All this is so scary. I don't want to jump into that." And then one by one, I've heard people like yourself come on and share, "Yeah, I had no idea how to do this thing." And then I was curious, so I just tried it. And you know, now I've shipped this amazing shader interaction to one of the more popular mobile apps in all of tech, you know, and it's so inspiring, especially given your background where, man, even just down to the last six months, you can just do things.

是的，你真的可以。我认为总是以好奇心为先导，正是这一点促成了那个时刻。所以在之前的几年里，我一直都在摆弄工具、做实验，学习 3D 的工作原理，学习动画和运动的工作原理。所有这些最终都导向了语音模式中的着色器，并且能够按照我们想要的方式进行艺术指导，因为你已经在其他工具中做过类似的事情，所以你能更好地描述你想要的东西。而且，是的，我认为这正好说明了好奇心是如何自我累积的，你也应该始终以好奇心为先导，因为它会带你到你还不知道的新地方。

Yeah, you really can. I think there's this element of always leading with curiosity and I think that's what sort of led up to that moment. So for for several years prior, always just playing with tools and experimenting, learning how 3D works, learning how animation and motion works. That all really led up to to that moment of shaders in voice mode and being able to art direct them exactly how we wanted because you've already been in other tools doing similar things so you could better describe what you wanted. And yeah, I think it just goes to show like how much curiosity compounds on itself and you should always lead with that too because it'll bring you to to new places you don't even know about yet.

我非常相信视频在解释设计师想法方面的力量。所以当需要反馈时，我会在 Slack 里丢一个 Loom 链接，再丢一个 Figma 原型链接，然后反馈散得到处都是，简直一团糟。所以我正在构建我一直希望存在的产品，它叫 Inflight。你可以把它看作一个异步评审工具。它让你可以轻松地分享视频演示以及交互原型或你设计的任何东西，然后 AI 会采访你团队的人，获取你需要的反馈，并把所有内容整理成一个漂亮的洞察页面。目前我只对 Dive Club 的听众开放访问。所以如果你想成为第一批使用 Inflight 的人，请访问 dive.comclub/inflight 来抢占名额。我喜欢把 AI 看作一个放大器，对吧？就像如果你在某个领域有一点点知识的种子，它突然就变得价值 10 倍。

I'm a big believer in the power of video to explain my thinking as a designer. So when it's time to give feedback, I'll drop a Loom link in Slack and another link to a Figma prototype and then feedback will be scattered everywhere and I mean it's a mess. So I'm building the product that I've always wanted to exist and it's called Inflight. You can kind of think of it like an Async Crit. It's an easy way to share a video walkthrough along with an interactive prototype or whatever you're designing and then AI interviews the people on your team to get you the feedback that you need and organizes everything for you in a beautiful insights page. So right now I'm only giving access to dive club listeners. So if you want to be one of the first to use inflight, head to dive.comclub/inflight to claim your spot. I like to think of AI as an amplifier, right? Like if you have the tiniest little seed of knowhow in a given area, all of a sudden it's 10x more valuable.

所以当你用这种视角来看时，很容易得出一个结论：尤其是在初创公司，通才设计师的价值是不可思议的，对吧？你可以通过拥有与传统产品设计框框相关的周边兴趣，以及好奇心和尝试不同工具的意愿，看看 AI 如何帮助你在所有这些不同方向上扩展，从而以多种方式做出贡献。

And so when you use that kind of lens, it's very easy to arrive at this conclusion that especially at startups, the value of a generalist designer is unreal, right? Like you can contribute in so many different ways just by having tangential interests to what we think of as the traditional box of product design. and then a curiosity and willingness to try out these different tools and see how AI can help you expand in all of these different directions.

百分之百同意。是的，完全正确。这是一个疯狂的时代。

100%. Yeah, it's exactly right. It's a crazy time.

你在 Perplexity 能够做出的贡献范围有多广？

How wide is the spectrum of contributions that you're able to make at Perplexity?

这是我喜欢 Perplexity 的一点：它非常适合通才，你可以以多种不同的方式灵活发挥。我们很多设计师都在写代码。有些更偏向设计与工程结合，有些则更偏向产品与品牌结合。所以团队中有着非常酷的不同类型的设计师。我认为我们有很大的灵活性去做自己想做的事，以及对于手头的任何任务来说合理的事。对我来说，这种能力不仅体现在产品工作上，还体现在帮助将品牌世界融入其中，并为品牌方面的工作做出贡献。比如，当我们为移动应用发布功能时，我经常负责这些功能的营销、展示视频和演示，通过动画和呈现方式，以我们认为最能传达这些功能的方式展示它们。所以你可以跨界进入品牌和营销领域，反过来也可以将更多品牌元素融入产品。甚至理解品牌在产品中的意义以及如何思考它。我认为这对我来说也很有启发性。

That's one of the things I love about Perplexity is it lends itself really well to generalists and you can flex in so many different ways. A lot of our designers are writing code. Some of them are more design and engineering built. Some of them are more product and brand built. So you have these really kind of cool different spectrums of designers across the team. And I think we have a lot of flexibility to sort of do what we want and what makes sense for any given task we're working on. And I think how that's showed up for me and one thing I've really appreciated is that ability to not only work on product but also help bridge the brand world into that and contribute to some of the brand side of the work too. So for example, we're launching features for the mobile app like I would often end up doing the marketing for those and the show reels and demos of those features and being able to animate and present those in the ways that we felt best communicated those features. So you're able to sort of cross into brand and marketing and then the other way around is working more brand into product. Even understanding what it means to have brand in your product and how you think about it. I think a lot of that has sort of been eye opening for me too.

我们能深入谈谈，对你来说，成功将 Perplexity 品牌融入移动应用具体意味着什么吗？

Can we go deeper on what it means to you to successfully integrate the Perplexity brand into the mobile app specifically?

是的，当谈到将品牌融入产品时，我的看法是，我经常将品牌和产品视为产品如何以非显而易见的方式兑现承诺。当然，有视觉识别、产品调性和产品中的视觉品牌元素，但归根结底，产品如何传递其价值？如何兑现这些承诺？我认为 Perplexity 通过几种不同的方式做到了这一点。此外，品牌本身非常以好奇心为驱动。那么，如何激发更多好奇心？如何将这种感觉带入产品？你可以通过后续问题看到这一点，这是 Perplexity 的另一个核心理念——所有那些让你深入探索的后续问题。它也在某种程度上训练你提出更好的问题，问出你从未想过的问题。这些都是我认为品牌与产品结合的不同方式，可能不那么明显，或者与你可能想到的仅仅视觉层面不同。

Yeah, when it comes to implementing brand into product, the way I see it is I think of brand and product a lot just about how the product keeps its promises in non-obvious ways. So, of course, there's visual identity and there's tone of the product and there's visual branding elements in the product, but at the end of the day, like how is the product delivering on its value? How is it keeping those promises? And I think Perplexity does that in a few different ways. There's also just the brand is very curiosity-driven. So, how do you evoke more curiosity? How do you bring that feeling into the product? And you could see that via the follow-ups, which was another idea that was founded on in Perplexity was all of the follow-up questions where you could just go down rabbit holes. And it's also kind of training you to ask better questions to ask questions that you hadn't thought of. These are all the different ways I think of working brand and product that maybe not be so obvious or like different from just a visual layer that you might think of.

我现在正在处理很多类似的产品挑战。其中一个我真正在纠结的问题是后续问题的价值，比如基于文本的后续问题，以及如何在语音模式下重现那种感觉，因为能够获得一系列选项并感叹“哇，我没想到那个”是非常强大的。同时拓宽 AI 能给出的回应范围。这里面有很多有趣的设计挑战。这是一件非常新颖的事情，为了做到这一点，我们也在思考如何在语音上下文中实现它——比如我们是利用视觉界面来做快速操作，还是将其融入提示本身，在最后提示你其他问题，但这也存在权衡，因为你希望它感觉非常快速和简洁。我认为我们还没有完全在语音中搞定它。

I'm working on a lot of similar product challenges right now. And one of the things I'm really wrestling with is the value of the follow-up questions, like the text-based follow-up questions, and how to recreate some of that feeling in voice mode, because it's so powerful being able to get the array of options and be like, "Wow, I wouldn't have thought about that." And widening the array of responses that an AI can give. There's a lot of fun design challenges in there. It's been such a novel thing and to do it we've thought about how to do it in a voice context as well like do we use the visual surface for that as like quick actions or do we actually work it into the prompt itself so it would prompt you at the end with like other questions but then there's also trade-offs with that because you want it to feel very fast and concise. I don't think we've totally nailed it in voice yet.

这很棘手。就像昨天我花了好几个小时和 GPT 来回模拟，试图思考如何为每个回应创建一个微妙的对话分支——比如这是你深入探索的路径，这是你转向其他方向的出口。在语音交互中，你能多快地在对话中给出两条路径？我不知道。我那一刻想，天哪，我已经做了好几个小时了。这感觉像设计。它看起来完全不像我习惯的那样，但我仍然在扮演设计师的角色，你知道吗？只是非常非常不同。

It's tricky. Like I was literally yesterday spending hours with GPT going back on these simulations and trying to think about almost creating a subtle conversational fork with each response where it's like here's your path to go deeper and then here's your exit ramp if you want to go somewhere else. And how quickly can you give two paths conversationally in a voice interaction? I don't know. I had this moment of, man, I've been doing this for hours. It feels like design. It looks nothing like what I'm used to, but I'm still acting as the role of a designer, you know? It's just very very different.

这仍然是设计，通过提示和塑造。这是一种新的技能组合，也是设计师可以利用的新途径。

It's still design through prompting and like shaping. It's sort of a new skill set and a new avenue for designers to leverage.

我们聊了很多关于语音的话题。你在 Perplexity 已经一年半了。事情发展得飞快。还有什么你参与或发布的工作是你特别自豪、我们可以强调的吗？

We've talked a lot about voice. You've been at Perplexity for a year and a half. Things are moving at a million miles an hour. Is there anything else that you've had a hand in or ship that you're particularly proud of that we can highlight?

最近的一些工作是我们重新设计了发现信息流，与刚加入 Perplexity 时做的不同。我们重新设计了主信息流，也重新设计了页面。对比起来很酷，因为我们看到由于这些设计选择，信息流的留存和参与度有了很大提升。所以这也是一个很好的学习时刻，了解哪些有效，哪些无效。

Some of the most recent work is we also redesigned the discover feed again from what we did right when I joined Perplexity. So, we redesigned the main feed. We also redesigned the pages. And it's been cool to compare because we've seen a lot better retention and engagement on the feed with some of those design choices that we made. So it's also been a really good learning moment on like what's been working and what's not.

你能描述一下其中一些变化，让我们了解背后的思考吗？

Are you able to describe some of those changes where we can get a sense of what the underlying thinking is?

我们这样做有几个不同的原因。一个原因是我们的内容量大幅增加。现在信息流中能展示更多故事。以前这很大程度上是手动过程，非常手工化、精心策划，每天只有有限的新故事可以互动。所以这就是为什么我们采用了更沉浸式的信息流格式，一次只展示一个故事。但现在内容量已经增长了 10 倍。我们需要一个稍微密集一点的信息流，让消费路径更容易一些。所以你在滚动信息流时可能会一次看到两三个故事。这样你可以更快地消费，也更快地找到相关或有趣的内容。这是其中一方面。另一方面是个性化和排序。我们做了很多工作来更好地个性化你的信息流，而密度也有助于此，因为我们可以确信，如果你一次看到几个故事，其中可能有一个是好的、相关的，你可能会与之互动；而如果我们只展示一个故事，那它必须非常精准，对你来说很好，标准更高。

We wanted to do it for a few different reasons. One reason being that our volume of content has picked up a lot. We're able to surface a lot more stories in feed now. Before it was very much a manual process. It was very handcrafted, hand curated and you only had so many stories, fresh stories a day to engage with. So that's why we did more of an immersive feed format that was like a single story at a time. But now the volume has, you know, 10xed from there. And we needed just a little more dense of a feed to make that consumption path a little bit easier. So you might get two or three stories at a time as you're scrolling the feed. So you're able to just consume much faster and find something relevant or find something interesting faster, too. So there was that side of things. There was also the personalization and the ranking side of things. So doing a lot of work to better personalize your feeds and the density also helps with that because we can be confident that if you see, you know, a few stories at a time, one of them's probably good and relevant and you'll probably engage with it versus if we want to make the bet that we're just going to show you one story at a time, like it better be spot-on, you know, good for you, higher bar.

是的，没错。所以有个性化这个方面，它还在不断改进。另外还有页面本身和故事，以及重新格式化这些故事，让用户对故事有更多控制。所以故事详情在格式上也进行了重新设计。延续了简洁和密集的原则，我们把故事稍微缩短了一些，让它们更容易阅读。我们还提供了工具来操作这些故事。你可以设置偏好：想要完整的故事报告，还是只需要摘要要点？然后我们会把这个格式应用到所有其他故事上。我们有一个小开关，可以选择摘要或报告。这样你就能更快地消费内容。我们看到那个页面重新设计后效果很好。

Yeah. Exactly. So there's the personalization aspect which is still getting better and better. And then there's also the pages themselves and the stories and reformatting those stories and allowing more control over those stories. So the story details also got a redesign in format. So carrying over that principle of conciseness and density to the content. We shortened up the stories a little bit just so they're easier to consume. And we've also given you tools to manipulate those stories. You can set a preference. Do you want more of like a full report of the story or do you just want the summary bullets? And then we'll carry that format through for all the rest of your stories. We have a little toggle where it's like summary or report. So you're again able to consume much faster. And we've seen a really good lift on that page redesign also.

这很酷。我得试试看。我在 Particle News 里看到过类似的模式。不知道你有没有……

That's cool. I have to play with that. I saw a similar pattern in Particle News. I don't know if you've...

是的，我就是从那里记下来的，心想这真的很酷，因为根据故事的不同，我肯定希望那个初始的 TL;DR 有非常不同的形式。

Yeah, that was where I saved that to my notion where I was like, this is really cool because depending on the story. I definitely want to have that initial TL;DR take a very different shape or form.

是的。这也影响了我对 Inflight 的很多思考。光是听你说“我们缩短了故事”，这虽然是个小细节，但本身就有点疯狂，对吧？历史上，我们总是围绕用户生成内容来设计，你得到什么就是什么。现在有能力真正定制输出，以符合设计师的目标，这完全不同了。

Yeah. And it's influenced even a lot of how I'm thinking about things for Inflight, too. And even just hearing you say something like we shortened the story. It's such a small detail, but that alone is kind of crazy, right? Like historically, we've always designed around user-generated content and you get what you get, you know, and now having the ability to really tailor this output in a way that meets your goals as the designer is totally different.

是的，确实如此。这也完全是为了让用户掌控，让他们选择内容如何呈现给他们。这只是朝着那个方向迈出的一步，与个性化一起。但所有这些都归结到同一个范畴。

Yeah, it really is. And it's also all about just putting users in control and letting them choose how they want this content to surface to them. And this is just one step into that direction too, along with personalization. But it all kind of wraps into the same bucket.

我很好奇一件事，尤其是从所有这些动态内容和个性化的角度来看。你加入时只有四个设计师，产品套件较小，而现在增长如此之大，你们有了独立的 Android 移动应用、Mac 应用，还刚发布了 Comet。我想了解你们有多重视，或者如何看待产品套件之间的关系，涉及系统设计、视觉语言、设计师之间的协作重叠。你能谈谈 Perplexity 目前在扩展产品表面积方面的策略吗？

Something I'm curious about, especially through the lens of having all this dynamic content and personalization, but when you joined there were only four designers, a smaller product suite, and there's just been so much growth where now you have the separate Android mobile app and Mac app and you just released Comet. So I want to get a sense of how much you all prioritize or how you think about the relationship between that product suite as it relates to systems design, visual languages, overlap between designers collaborating. Can you talk a little bit about what's the strategy at Perplexity right now as you all are growing the product surface area?

是的，在扩展表面积和同时壮大设计团队方面，我们有更多的贡献者。它仍然相当灵活和松散，我们的心态基本上是：有一致的设计基础，但你可以随意使用它们，塑造你正在做的东西，只要基础一致。这是整个产品套件中最重要的事情。但你仍然有很大的灵活性。而且我们也被授权——我的意思是每天、每周我们都会遇到新的挑战、新的用例或新的需求，这可能需要不同的视觉语言或不同的视觉处理，然后融入基础中，我们对此并不严格。如果有意义并且我们需要，我们就会把它加进去。令人惊讶的是，这也能规模化，并可能影响正在进行的品牌工作。然后，对于不同的产品表面，我们希望它们感觉独特且有差异化，让你有理由选择其中一个而不是另一个。我们不希望只有一个产品在所有方面做同样的事情，或者你得到相同的界面。所以 Android Assistant 有点像操作系统接管，它非常依赖于你正在做的事情，你只是与这些小部件交互。移动版应用，我们实际上正在重新设计，目前它在导航和某些交互方面感觉像是网页产品的反映，所以实际上相当接近和一致。但新的重新设计，我认为，感觉更加专注和更有主见，重新优先排序并重新强调了一些导航元素，真正提炼出你想要使用的核心移动用例。这就是我所说的审视我们的产品套件，让它们有意地专注于某些任务。你会看到这在 Mac OS 的一些新想法中体现出来。网页应用当然有自己的表面。我们有 Comet，它也有新的表面，比如 Comet 助手。但我们真的想推动每个表面的专注性和意图性，以最好地服务于你所在的上下文或该平台所需的用例。

Yeah, in terms of growing our surface area and growing the design team at the same time. We have many more contributors. It's still fairly flexible and loose and our mindset is pretty much like let's have consistent design foundations, but you can kind of use those however you want and shape whatever you're working on as long as it's foundationally consistent. That's what matters most across the product suite. But you still have a lot of flexibility within that. And we're also empowered to, I mean every day every week we're coming across new challenges or new use cases or new needs for something and that could afford some different visual language or some different visual treatment or something to roll into the foundations and we're not strict about it. If it makes sense and if we need it, we'll roll that in. And it's surprising how that can scale too and inform maybe some of the brand work that's happening. And then in terms of the different product surfaces, we want them to feel unique and differentiated and like you have a reason to use one over the other. We don't just want one product that's going to do the same thing across the board or you might get like the same interface. So Android Assistant is sort of this OS takeover. It's very contextual to what you're doing. You just kind of have these little widgets that you're interacting with. The mobile version of the app, which we're actually working on a redesign right now, where today it kind of feels like a reflection of the web product in terms of navigation, in terms of some of the interaction. So that actually is pretty close and consistent. But the new redesign, I would say, feels much more focused and much more opinionated and kind of re-prioritizes and re-emphasizes some of those navigation elements to really distill it down to the core mobile use case that you want to use it for. So that's kind of what I mean by looking at our suite of products and making them feel intentionally focused on certain tasks. And you'll see that show up in some newer ideas for Mac OS. Web app of course is its own surface. We have Comet that also has new surfaces with like the Comet assistant. But we really want to drive focus and intentionality to each of these surfaces to best serve the context you're in or the use case that you need for that platform.

而且我想作为设计师，这样可能也更有趣。获得更多创作自由，可以展示品牌实力，看看会发生什么。

And I'd imagine it's probably more fun as a designer that way, too. Get a little bit more creative freedom, get to flex the brand muscles and see what happens.

是的，100%。

Yeah, 100%.

我们聊了很多。就我个人而言，我非常受你作为自学者的旅程启发，你显然充满好奇心，并且有自主权去尝试和实现事情。所以也许在个人层面上，在我让你走之前，作为一个喜欢自学并且对世界现状和未来方向有所了解的人，你如何看待作为设计师想要投资的领域，以及未来提升自己技能的方式？

We've covered a lot of ground. Personally, I'm very inspired by your journey as a self-learner, someone who is very evidently curious and has the agency to just go try things and make things happen. So maybe even on a personal level before I let you go, as someone who enjoys self-learning and is also kind of plugged into the state of the world and where things are heading a little bit. So how do you think about the areas that you want to invest in as a designer and the ways that you want to level up your own skill set moving forward?

我认为提升的方式就是把握事物的发展方向，然后据此塑造自己。所以对我个人来说，哇，编码现在变得更容易了，有了像 Cursor 这样的工具。那是我真正需要学习和掌握的工具，因为我看到了它的价值和优势，并且亲身经历过。所以对我来说，那绝对是我在提升技能的领域。我想贡献更多代码，用代码构建更多东西。看到我们现在的工作方式以及这样能快多少，这显然是我看到的行业脉搏之一，就是扩展你的工具集。老实说，对于产品设计师来说，这可能是最主要的。如果你还不了解它，现在没有理由不投入其中。

The way I see leveling up is just trying to get a pulse on where things are going and sort of shape yourself from there. So for me personally, it was like, wow, coding has become much more accessible now with tools like Cursor. And that's a tool I really need to pick up and learn because I see the value and the advantages and I've experienced it firsthand. So for me personally, that's an area I'm definitely upskilling in. Like I want to contribute more code. I want to build in more code. Just seeing the way we work now and how much faster we can work that way. That's clearly one pulse on the industry I see is just expanding your tool set there. That's probably the dominant one to be honest as a product designer. If you don't already know it, like there's no reason not to jump into that now.

而且我认为这话从你嘴里说出来很有分量，因为你来自比我更偏艺术的一端。你知道，我一直更像一个技术系统设计师。

And I think it carries a lot of weight coming from you because you are coming from a more artistic end of the spectrum than I am. You know, like I've always been more of a technical systems designer.

我多年前上过一个编程训练营，大部分都忘了，但确实上过。你知道，我大体上对那个感兴趣。

I went through a coding boot camp many years ago. Forgot most of it, but I did it. You know, like I'm generally interested in that.

嗯。

Yeah.

而我呢，什么都不会画，你知道，我刚在 Jitter 里开始做动画。我的艺术技能水平很低，但你在那个领域很突出。所以，你带着那样的背景，却说“我投资的是手艺中更技术性的一面”，这话在我心里很有分量。

Whereas I can't draw anything and I, you know, I've just started animating in Jitter. I have such a low level of artistic skills, you know, but like you spike in that area. And so for you to come with that background and still say the thing that I'm investing in is more of the technical side of the craft carries a lot of weight in my mind.

我是说，我也可以这样来表述：你也需要评估自己，评估你处于什么位置。所以，如果你偏技术，可能不太艺术或不太视觉，为什么不开始涉足那个领域，看看会打开什么机会，你能从中学到什么。所以，我认为需要一些自我反思来理解你想往哪里走。因为就我而言，我至今做的大量工作都是视觉、品牌、动态，与产品设计相关，而编程对我来说是个异类。所以，那是一个明确的方向，我想进一步深入。

I mean, I could also frame it this way in that like you also need to assess yourself and assess where you're at too. So if you are like heavily technical, maybe less artistic or less visual, why not just start dipping into that and see what opportunities that opens and what that could learn or what you could learn from that. So, I think there needs to be some self-reflection to understand like where you want to go. Because in my position, a lot of the work I've done up to date has been visual, brand, motion tied in with product design and coding was kind of the outlier for me. So, that was a clear thing like I wanted to dig in further.

我每年都努力做到这一点。我告诉自己，我想以某种方式重塑自己，或者学点新东西。所以有一年是动画和 After Effects，另一年是 Cinema 4D 和 3D。嗯，今年是 Cursor，是编程，就是那个专注点，花一年时间深入钻研，所有这些经验和知识真的会随着时间不断累积。

I try to make a point of this every year. I tell myself I want to kind of reinvent myself in some way or learn something new. So one year it was animation and After Effects, another year it was Cinema 4D and 3D. Uh this year it's Cursor, it's coding, it's that focus just to spend a year going deep on and all of that experience and knowledge really compounds on itself over time.

嗯，Gunnar，我深受启发。感谢你来做客，分享更多关于你自己和你的经历。我希望听众们受到鼓励，迈出新步伐，尝试新事物，在职业生涯中多一点通才心态，因为这很有效，而且看起来很棒。所以，感谢你今天来做客并分享。

Well, Gunnar, I'm very inspired. Thank you for coming on and sharing a little bit more about who you are and your journey. And I hope that people listening are encouraged to take new steps and try new things and take a little bit more of a generalist mentality to their career because it's working and it's cool to see. So, I appreciate you coming on and sharing with us today.

哦，是的。我很感激，老兄。谢谢。

Oh, yeah. I appreciate it, man. Thank you.

在你走之前，我想花一分钟给你介绍我最喜欢的产品，因为我经常被问到我的技术栈是什么。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 是我为设计做动画的工具。Lovable 是我用代码实现想法的工具。Mobin 是我找设计灵感的地方。Paper 是我像创意人一样设计的工具。而 Raycast 是我每一步的快捷键。现在，我亲手挑选了这些公司，这样我才能全职做这些节目。所以，到目前为止，支持这个节目的第一方式就是去看看它们。你可以在 dive.comclub/partners 找到完整列表。

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. Mobin 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.comclub/partners.