Linear's Karri Saarinen: design-in-code tools risk killing conceptual exploration — the real source of innovation

our next guest. Let's bring in Ki from Linear. Uh, welcome to the stream. He's in the roosting room. How are you doing, Ky? Good to see you.

What's happening?

I'm doing well. How are you guys?

We're doing fantastic. We're in the Christmas spirit. Obviously, we're dressed up like Santa Claus. Both Santa Claus,

if you didn't notice.

If you didn't notice. [laughter]

Yeah. I mean, it's it's going to look similar what you usually look like. So, notice it. When do you think is the right time for companies to start getting into the Christmas spirit? Because Amazon was running kind of Christmas themed ads back in October. That felt a little early for us. [sighs and gasps]

Yeah. I mean like I I don't really have like opinion on this on the company level. Like we don't do anything about it. But um I think personally like I I believe in the

in the like after Thanksgiving it's okay to start it. Um, so you can get a Christmas tree after Thanksgiving

after Thanksg Let's hear it for another respectctor of after I said the exact same thing earlier.

Everyone says this, but uh there's clearly been a uh a sort of erosion and and every day every year, one day earlier, and now they're running Christmas ads in October. It's an outrage. Um but anyway, um we're not here to talk about Christmas anymore. We're here to talk about design. We're here to talk about um your essay. Design is search. I would love for you to just kick us off with like the inspiration behind the essay, why you felt the need to set the record straight on this particular topic, and then we can go into some of the actual uh the actual thesis and hopefully bat around some different ideas around it.

Yeah. Um yeah, I think this whole like debate on Twitter or on X started like on on Friday. I was just sitting on the couch and and in the midnight and thinking about this that this is probably like I don't know the 10th time I've seen a company launch this like design to code tools or code in design or like design and code kind of tools and

I I think it's like people always kind of see that that's kind of like the final form of design like now we can finally design in the actual medium like actually in the code

and I think it's there's definitely like some benefits to that and like it it is a great tool but I always think like design is Not just the tools, it's it's like broader than the tools. And in a designer's toolkit, you always have many tools and then depending what you're designing and like what this like what is the environment like who we're designing for, what's the domain, what's the product, you you use different tools. So I think like it's not like that clearcut that like like designing in code is the best way to do design. So I kind of like wanted to like highlight that there's some like challenges with this. And what I started with was that the I think that what also happened to me in the past like when I was designing and building front ends for apps. Um I [clears throat] what I what often like start feeling happening to me is that like I I become less idealistic or I become more conservative because like I know that like whatever I design I also have to build. So there's like already like you start hesitate yourself or like you don't believe um you can actually build certain things because like you you start to kind of collapse the design and build more too much. Um but I think like also like if you are designing against an existing code base I think it's also introduces like additional bias where it's like the code base is like is this kind of collection of um decisions and architecture and layers and and and product features that have been built over the years or over the months and you are now kind of like trying to fit something new into that. So like the natural tendency there is to like fit something that fits and like kind of like any kind of new or alien idea can kind of like feel like it doesn't fit and doesn't feel like a good idea. So like you are too much reflecting the design against the state of the world versus like thinking about design more in abstract like what the world could be.

It's like an art artificial constraint.

Yeah. I mean it's it's not Yeah. I don't know if it's an artificial constraint but it's a constraints what arrives too early. So like eventually any design has to be like manufactured or coded or something like made it real like design is not real. It's like it's an idea or like a concept or like it's it's something but it's it's not something people use. So I think it's eventually you have to deal with the constraints similar like architects has to deal with their constraints like how does the building like uh how can we build this building? How can we staying under budget or how do we make sure the materials how much usable land do we have that kind of

Yeah. But it's like I think like lot of science architects still don't start from the constraints like they might acknowledge the constraints but then they go to start sketching something what they want to see.

Yeah.

So I just want wanted to highlight that like every tool has certain certain kind of strengths and weaknesses weaknesses and like if you're not mindful you kind of like they can influence you your thinking a lot. So what I worry about is that like if we start thinking design is about coding then I think we lose something like we lose that like space to think more broadly or kind of without those constraints and we jump into like let's build this solution and that was kind of like the I think like some of people were pushing back on this essay or this ideas I was I was talking about that well now the design can be better because we can build it better and the quality can be better but again this kind of proves proves my point. It's like I don't think that uh I think construction and execution is different from designing and I actually think that while it can help to construct the design better like does it really make the design better? So I'm like still like looking for that like how do these tools actually make the design process better? Like how do we make better things with these tools versus like we just execute them faster?

Yeah. Is your is your view that like doing design within code is something that should uh effectively happen at a refinement stage where maybe you have a process of like planning how you want to build a product or a feature and then you know you think really deeply about how you want to do it. You make a bunch of decisions based on your intuition and what you know about the customers and how the thing should work and then you build it and then at a certain point you're able to make like more kind of like microlevel changes to it when you've already you've already done all the work of the kind of like heavy planning and and being really thoughtful about something versus I don't know I I would be I'd be worried if if um uh you know somebody was like um building me a house and they just came to the land and they threw up like a box like a frame and then they started just kind of like okay we're going to add a door over here and oh we actually have to add like some pipes under here so let's like dig under there uh and like you you want when you're trying to build things that are high quality and great I think like that uh that like pre-planning process and doing things in the right order matters a lot. So I feel like with these new tools you have to be thinking about where do you actually like slot in that where do you slot into the tool to the overall process.

Yeah. Yeah. I think like I think these new tools can can help to blend this like so I think in the past I think there's more like this like I don't know the way things operate is much more much more that there's a design and then there's a handoff then there's a engineering and then nothing crosses this border and I think that's a bad way to do as well and I think it there should be this blended mode where I like to think that design is kind of essentially like two phases like you have a conceptual design so you're kind of like trying to understand the problem and then you're trying to explore multiple directions how to solve that problem and like I always tell my designers to do that like even if you know the solution you should still explore something else because like maybe you find something better or like sometimes you should even explore bad ideas because sometimes the bad ideas show sometimes bad ideas actually turn out to be good ideas and then sometimes bad ideas can show why the the other ideas are good. So I think like there's this stage where what what my essay was about was like design is a search. So you you're searching for the direction and like you're searching for the concept. You're kind of like removed from the execution stage of actually making it work. So the conceptual design is like it's kind of like cars car industry has a concept cars like those cars are not in production yet and like they never will be but it's an idea what the car could be and like car industry never like seems like they never take anything from those concept cars [laughter] but but they could

we took the handle we took the handle from the passenger door.

Yeah. we made this crazy Gullwing super fast crazy crazy car and it's like yeah the lights

so that that's kind of like the idea is like there's a concept that is cool and we should try to execute it but like I don't know what happens in the car industry they just kind of like can't do it but I think like in a software you can do it and then there's the execution design where you like how do we actually make this work well in this context of the application or fix the like fix the problems we we didn't think about in a concept like the edge cases or performance or just like we get user feedback, we make some fixes. So at that stage I think the coding tools can be really useful because like you are iterating on the you're refining the the the kind of like the design but you had first this other stage where you kind of like thought about it like more like I don't know not exactly first principles way but like you started from like kind of like a pure problem or an idea and then you like created some concept that you think is worth like um executing on.

Yeah. Uh, how do you think about like it feels like so much of what AI is great is maybe I mean AI is pretty good at like search but I still feel like we probably agree that there's elements that are lacking around inspiration and and and design is certainly not a solved problem from an AI perspective but uh a lot of times translation is more or less solved. So uh what stuck out to me was in your essay you you noted that a lot of architects will start by sketching with like pencil and paper and we are in the area the era of like you could sketch it and maybe you have to come up with the brilliant idea or the brilliant beautiful design but then you can just take a picture of that sometimes and get a floor plan out of it potentially. Um, but do you think that something's lost there or do you think that like text like like are these all going to be lossy and so you want to avoid like text to image to code uh and you want to be able to like dip in at every single level? Like how do you think about the different translation modes because some of them are really helpful like at the end of the day some of them are just autocomplete and that's amazing. Mhm. Yeah. I I feel like there could be always something lost. Like I think like when you actually if you draw something and then you try to like draw it again for example, you you kind of like even if you just took the sketch and then draw it again on a paper, you probably like start noticing different things like you start thinking about the problem like slightly like

different way. So I think like what the AI is kind of like always doing is kind of like it's shortcutting you to the output without actually you yourself going through the process. So I think like something always gets lost there. I think again and I don't think it's like necessarily a bad thing. You can just like take a sketch and generate an idea and like you see it in a different way and that can be good too but I think there's that that problem and yeah like some of the comments I got that's like well now the AI is so good so you don't even have to code. So it's like you just describe what you want and the AI kind of like creates it for you. But then now you're kind of like adding another another layer of influence there. It's like the AI is like kind of influencing the output like and the LLMs are kind of like

statistic models whereas they're like always like looking for the correct like kind of like statistically correct answer. They're not looking for the like necessarily the unique answer or your best best answer for you personally or for this like context or something. So like you there there can be this danger of like everyone starts to average towards the means or the the the the kind of like the the common standard patterns or standard practices. Um again like I think it's fine to use AI but it's just like kind of like you have to understand like what are you giving out or what is the tradeoff

and like if we were trying to do something about it or are you okay with it the mdash of code? Well, one one interesting example earlier in the show we were talking about uh where's Waldo. We we use like Waldo bench. So, we'll test like a new image model by just making like a Where's Waldo illustration

and like none of the models can do it yet. It's like really complicated. They can make it visually look like it's perfect, at least from far away. And then you zoom in and it's like chaotic and there's no story happening around the page. And maybe Waldo is just not even on the page. Like it's the ultimate where's Waldo? He just bounced. Um, but we looked back at the history and the the original illustrators, this guy Martin, and he would spend like eight weeks on to make like a full spread, right? And you could imagine uh I mean part of that was just like hand illustrating everything, but it's also like planning out like what is the story line and what is the experience of like being in this page and like what is the experience of finding Waldo, right? And I feel like that is the that is still the thing that is ultimately going to give products edge and make products have that craft is when you know that a team that's building software could have oneshotted like a dashboard and got it like to to to the point where it's like it's functionally a dashboard. Like we can functionally make a where's Waldo illustration that you can go on the page and search for where's Waldo, but is it like an enjoyable is it an enjoyable experience? Like right now we're not seeing AI that like maybe that you could find the right prompts, but right now you're not doing you're not oneshotting an illustration like this and and it is still something you have to put an immense amount of like thought and planning into and and uh try things and and revert them things like that.

Yeah. And I think there's different modes like I think you can use AI like creative ways of like generating a lot of output or different ideas but I think it's it's like yeah there's always like you now have this influence of the AI model like what is it what how it does how it creates those outputs and like decides things and then you have the tool that is doing like the code base versus like you're kind of directly working on a sketch or a screen where there's like very little like influence like what what can be do how the design can be the the other reason like I like why I even like talked about this like like I didn't I don't I don't have any tool to sell here like I don't have a design tool to sell but it's it's more like I worry about the if we start to like kind of see design as coding then or like start to see design as construction that we like already like there's like the incentives to ship things fast is so strong that we start to see that oh like now we don't need to design because we can just build it and we can designers are going to like design and build it and then I think we start losing the the other phases where can we just think about the problem like can we think about the concepts like how could this be work really nicely um and then like I think we it's like very like in like it's very exciting idea for companies or founders or someone like I don't have to think about design anymore it just like happens with the when we build And I think it's it's it's maybe like a good for the business, but I think it's just like not good for the world or like good to get like my my kind of like motivation with linear and like talking about some of this stuff is always that like can we just have nicer things in the world? Can we like spend more time on the craft and quality and and like value it again even if it's not like something that can be measured? And I think the Silicon Valley and like technology companies always had trouble with this. It's like it's hard to like value something that is not quantum available like easily. It's it's like

it's if there's no data for it then it's doesn't exist and like good design is something sometimes like it's it cannot be measured. So, so I think like this again like shipping speed and and like how many things you ship can be measured, but like if someone makes a design really good, it's like it's hard to like um it's hard to measure that and and then designing things or like sketching things can be seen as unproductive because it's not directly um contributing to the shipping speed or shipping more things. Mhm.

What's your philosophy on uh story story storytelling? Is is Matt your storyteller? Do you like that word?

Yeah. I mean I I like it because like I think I don't know what happened to marketing but I think marketing kind of like lost the lost the story [laughter] uh or lost the plot there at some point.

It became very quantitative. It like it became more mathematical like just buy the ads and run all the AB tests and all. It became more of a science and so the art needed a new brand maybe.

Yeah, that makes sense.

And like I would I would kind of like think like storytelling is kind of like I think it the same way as I think the brand. I think the brand and storytelling is maybe like the same thing. It's like so I and like to me brand is something just that you have a belief or like some kind of worldview and you like and you have some values or things you stand for like we always had that and planer and like that's kind of like from the place I always articulate my things. It's like I try to like keep true to that brand is like can we have nicer things or quality things or or something and so the brand or the storytelling is not never about the like again about the output. It's actually like it's like a deeper thing. It's like do you believe in something and then I think the storytelling of the brand is just like how do you express that belief? So anyone like people can't read your mind. So you have to like either tell stories or you have to

show visuals or videos or something. So how do you express your beliefs?

So the problem when people are looking for storytellers is that like again like do you have the story to tell like do you believe in anything and like is that If you believe in something, is it like engaging? Like do we people do people want to hear that story? So hiring a storyteller doesn't help if you don't have this like a good story or like engaging story to sell.

Yeah. And someone in the chat noticed you guys are hiring a writer, which is something I was pushing for. Like if I wanted to hire a storyteller, I would wanna personally I'd want to hire somebody that is very good at writing and wants to come in and write a lot and like enjoys the craft of writing and not somebody that wants like a like a pretty title that like sounds cool but then ultimately just wants to collaborate crossf functionally across the org or whatever it is. Um, last question I have. What are your thoughts on holiday gimmicks in products? I got into my car this morning and on my profile,

it had a a Santa hat on it and I I got I got some enjoyment out of it. It had some cobwebs for for Halloween, too.

Really?

Um but uh is that cheapen a product? It feels like something that uh

I don't I don't know which way you're going to go.

Yeah, I would think about it like again like from the point of brand is like does it like fit the brand? Like I think we we haven't done those things because we kind of like linear kind of the brand is more that it's like a serious tool for work and it's it's I think some of those kind of things can can kind of feel like off and like I don't know like for example it's always like these situations where I don't know like a company is using linear and then something goes down like their service goes down like their the whole company is like scrambling and then suddenly some Santa hat appears and everyone's like what is this thing it's like we we have like like our product or like customer base is on fire and like there's some Santa has flying around. [laughter] So it's it's like it can be distracting and so I think it's kind of depends on your brand and like how critical like infrastructure you are and where do you put these things. I think it's it can be fun thing for the team to do Easter eggs like we do Easter eggs that are hidden. So there there's like you can there's games you can there's an arcade you can enter in linear and then then there's like old arcade games you can play but you know you don't find them like you have to really find them to like you need to find the Easter egg to access it.

So there's like these kind of things that you make secrets but um but yeah I don't personally like the like putting the decorations on your face and in like professional tools.

Yeah, makes total sense. Well, uh, great having you on and, uh, yeah, we'll save the Santa hats for the podcasters.

Yeah,