An Interview with Figma CEO Dylan Field About Design and AI

Good morning,

This week’s Stratechery interview is with Figma co-founder and CEO Dylan Field. Field was a Thiel Fellow who dropped out of Brown in 2012 to start Figma. Figma was born of a technical breakthrough that leveraged WebGL to deliver powerful graphical capabilities in the browser; the browser made Figma collaborative, what I call the operating system of design.

Figma has had a fascinating road: the company accepted an acquisition offer from Adobe in 2022, but due to regulatory resistence the latter was forced to abandon the merger in late 2023. Figma instead IPO’d in 2025, and after skyrocketing to a valuation of $56.3 billion, has since crashed to a market cap of less than $10 billion, less than half of Adobe’s offer, thanks in large part to a market narrative that the company is an AI loser.

I talk to Field about all of this, including his background, Figma’s differentiation discovery process, and the nature of creativity versus design. We get into the AI question, which the market views as a headwind, but which Field sees as a tailwind. To that end, the occasion for this interview was Figma’s Config conference and Field’s keynote where he explained how Figma’s Canvas was the natural intersection between design and AI.

As a reminder, all Stratechery content, including interviews, is available as a podcast; click the link at the top of this email to add Stratechery to your podcast player.

On to the Interview:

An Interview with Figma CEO Dylan Field About Design and AI

This interview is lightly edited for clarity.

Background

Dylan Field, it feels like this interview has been in the works for years, but welcome to Stratechery.

DF: Thank you, appreciate you having me, and big fan.

Let’s start with your background. Where did you grow up, how did you become interested in technology? I always love these stories, especially the first time I talk to someone, and I think yours is a particularly interesting one. So give me the story.

DF: I grew up in Penngrove, California, which is near Petaluma in Sonoma County — but not Sonoma, it’s critical to make sure people know where Penngrove is. My mom was an elementary school teacher, my dad a respiratory therapist, both not especially tech-savvy, but my mom early on realized that a computer would be useful for me to stop bugging them with questions and bug the computer instead. So I was lucky enough to get a — I think it was a Compaq Presario — when I was like five the family got one, and then I proceeded to really hog it.

I’ve pretty much been interested in technology as far back as I can remember, I was very eager and excited to learn how to program, but didn’t necessarily have the ability to get my hands in a compiler for a while. It took until I got through some scholastic program, a BASIC compiler, to actually get properly started. I’ve also always had a, maybe not as much ability as I’d like, but a deep fascination with mathematics and just really everything in the world.

And so this is just a fascination with the technology — like, how does this thing actually work, and how can I make it do what I want?

DF: It was always more about product and design and about what technology will look like in the future and how to get there, rather than “I can really master the technology and have it under my control”, that was never really my vibe.

What were the sorts of things you imagined you wanted to make as a kid, when you have this computer you want to figure out?

DF: Walking around as a kid I was probably thinking less about the computer and more about, “Why can’t I teleport?”, or, on the flip side, going to SFO the first time and seeing they had these magical faucets where you put your hand in front and the water comes out and you didn’t have to touch anything — and I was a germaphobic kid — I’m like, “Why can’t the entire bathroom be automated?”, it’s just so obvious. Or, before I even learned how to properly read and write, “Why can’t I talk to the computer?”, stuff like that was more what I was excited by.

Are you encouraged or discouraged by the progression of bathroom technology over the years?

DF: Encouraged. Toto‘s wonderful.

Yes! It’s funny, because Toto is in the news because they make a certain sort of ceramic that’s used for AI stuff. I’m like, “Look, I’ve known about and been a Toto fan and supporter for many, many years”.

DF: (laughing) I didn’t know that. Well, the other critical design invention here, which is very underappreciated, if you’re leaving a bathroom and you can use your foot to pull open the door, that is an underappreciated progression.

Oh, there you go, that makes total sense, I can’t say I have that in my bathroom, but I do have a Toto Washlet toilet, they are well worth it — the only problem is you’ll be spoiled for life and won’t be able to live without it. So you end up at Brown — not what you’d think of as a technology school, it’s next door to RISD, which is a design school, so there’s an angle to where you ended up. What was the path to getting there, and the path to leaving as a Thiel Fellow?

DF: During high school I was probably a little overconfident, thought I could do anything and was beyond bright, and the world quickly proved me wrong, “Okay, there are people far smarter than you”. But due to that identity, I thought maybe MIT would be the place I want to go, then I toured MIT and it was a cloudy day, midterms, and I went, “No, this isn’t for me”, and looked at other spots.

One person I’d talked with a lot was Danah Boyd — I met her through O’Reilly Media — and she was a really brilliant, thoughtful person, and she said, “You’ve really got to think about Brown”, and I kept randomly meeting Brown grads as I was doing this East Coast college tour, very randomly, and they’d all sit me down for an hour and tell me, “You’ve got to apply to Brown, and if you get in, you’ve got to go”. I ended up applying to Olin and Brown on the East Coast out of ten schools I visited, I was thorough, I didn’t get into Olin, which I thought was my first choice at the time. And then Brown, I was very surprised but thrilled to get in.

What did you think you were going to study at that point?

DF: Computer science and math, I did formally declare that as my concentration, but I didn’t get as far on the math side as I would have liked — did more CS classes, and also took advantage of Brown’s amazing open curriculum, where you can go very broad, I had some incredible classes in areas that are not technical at all.

So where did the Thiel Fellowship come into the story?

DF: It was the fall semester of my junior year. I was aware of the Thiel Fellowship — I’d seen it online, thought it was kind of a weird idea, but interesting. I got introduced to it by Elizabeth Stark, who now is, I believe, leading Lightning, she introduced me to one of the Thiel Fellows at the time, Dale. It was this weird one where he was 25 minutes late to a 30-minute meeting at Starbucks — we met for five minutes, but then he just kept texting me, “You’ve got to apply to the Thiel Fellowship”, very similar to the Brown story.

I ended up applying after speaking with my now co-founder, Evan Wallace. Evan was the most brilliant person around — a year above me at Brown, my TA for multiple classes, and truly a genius, someone who’s also just fundamentally kind, humble, wonderful. I was like, “Man, I’ve done some internships now, there’s no one better to start a company with”, and if Evan were down for that instead of any number of jobs he can get when he graduates, I’d learn more from it than anything else — I can always go back to Brown, so I should at least explore it, and he surprisingly was down to explore it with me.

So I applied to the Thiel Fellowship with a drones idea — which I think now is best being done by BRINC. Evan was just not down for that direction, he was down for WebGL and graphics, and I was psyched by that too, that’s the direction we headed.

Tell me about the drones idea and the pivot to the WebGL angle, because it ties into the question I asked at the beginning — what were you pursuing? Was it the technology, or the end state? I think that’s an interesting through-line here.

DF: I’ve always been excited about a lot of things — creation, creativity, design, even before I knew what to call design, which was most of my life at that point, I’d only recently learned what the word “design” meant, despite having done a lot of design.

For me, I saw the act of starting a company was also about asking the question, “Why now?”, there are so many “Why now?” answers you can give, it can be societal change, cultural, technological, regulatory. But we were technologists at our core, so we made a big long list of all the technologies that were changing at the time and gradually crossed each one off, we came up with two finalists.

One was drones, this is the end of 2011, the other one was WebGL. I think we would have totally failed at drones anyway, it’s extremely hard. You look at Zipline, BRINC — these are amazing companies, and you really have to chew glass to get through that, we wanted to do something where we felt we had a technological edge and insight others did not.

WebGL and the Foundation of Figma

And what was the technical edge and insight about WebGL? This is obviously the foundation of Figma — you can do incredible graphical things in the browser, which to that point had all been on dedicated desktop applications. What was the insight that made you think this might be possible, even if it was just barely possible?

DF: To be clear, right after applying for the Thiel Fellowship with the drones idea, I ended up working at Flipboard as a design intern, using design programs all day long. We had this hammer with WebGL looking for a nail, we didn’t find the, “Let’s go build design environments and help designers”, for a while, it took a little bit.

What was exciting was that Evan had done a lot of early work that proved out that WebGL was way more capable than anyone else was thinking at the time. Other folks then were going, “WebGL is this weird toy that Mozilla is making, it’s probably not as important as just using your local, non-browser tech”.

Right, if you use an application that can actually leverage regular OpenGL and your GPU, why a browser?

DF: Exactly. The only other company that seemed on to it at the time was Onshape, actually. We looked around and went, “These guys get it”, and pretty much no one else did yet, no one took it seriously.

So due to Evan’s work, we started to really explore that and go, “How can we take tools that people expect to be desktop-bound and local, bring them to the browser, and do it collaboratively too?”. We were very inspired by Google Wave — rest in peace, it was a really cool product. I grew up in Google Docs, playing MMOs and stuff like that, so I think our frame of reference, even if we couldn’t articulate it then, was just different — obviously the browser enables all of that.

You viewed the browser as a first-class operating environment in a way that probably older people did not.

DF: Yeah, exactly. In the early days of Figma I’d say, “Just like Google Docs”, and a lot of people were like, “Yeah, well, I use Word — why would I use Google Docs?”, and I was like, “Well, I’ve only used Google Docs my entire life”. And then, “Well, I guess there was that time in middle school…”, and they’re going, “Wait, how young are you?”.

Well, let’s talk about what Figma is. I’ve written about Figma in contrast to Sketch, which is more of a single-player experience — this idea that Adobe left this huge window open for actually designing apps. Mobile apps come along in particular, an exploding market, actually placing all the screens, how it all flows together, they didn’t have a product for that. Sketch comes in and fills that gap, but it’s still an application on your computer, and you’re saving files that are v1, v2, v5000.

Figma, by virtue of being in the browser, got collaboration for free — it’s a multiplayer experience. When did that possibility become clear? You mention the collaboration aspects, but as I understand it, you were trying to get WebGL to work first, and then realized this is good for collaboration. Is that the right sequence, or did you have the benefit of being in the browser — meaning multiple people could work on something at the same time — all along?

DF: I would say from day zero, Evan and I were talking about it, and we were both trying to be very rational. On collaboration, we wanted to talk with users and see, “Do they need it?”, and basically everyone said, “Not only do we not need it, we don’t want it”.

Right, there was a lot of asking jockeys if they wanted cars.

DF: Well, I think it was more an identity thing of, “I’m a designer”, and there was a lot of agency influence on the design process at that time — this kind of grand reveal where you just work in the corner.

Oh yeah, you own it, it’s on your computer, you’re doing it, and then you go into the meeting and show it.

DF: No one sees it until it’s perfectly ready, then you show a few results, maybe give them three, the first two are kind of not what you want, but the third, “Oh, the contrast is so great”, and everyone goes with it. So that agency mindset and identity, as well as imposter syndrome, honestly, because design was just emerging from this phase where people saw it as, “Make it pretty”, versus, “Make it work”. This is a key element of how we build product, build software, do media and advertising, and people were just starting to appreciate it with all the Apple ethos of the time and great consumer products coming out.

So we had the insight from the start, but it took us a while. Eventually, as we built it out and started fully using Figma to build and design Figma, it was immediately clear there was no way we could launch without collaboration, because it just felt wrong. If you’re in Figma and I share a doc with you, a link, and you’re in it too, and I make a change and your browser force-reloads, and you make a change and my browser force-reloads, it sucks. So it was a, “We have to do this thing”, and it was not trivial at the time — it took quite a long time to build out. Evan was a key part of that, as he was with a lot of our foundational technology, it was a key condition for our launch in 2016.

Is it ironic that Apple sort of created the conditions for you in raising the stature of design and that being the controlling factor in development, even as their whole tech approach is counter to you, not really supporting WebGL, being all-in on applications? It’s kind of interesting.

DF: I don’t think Apple’s tech approach is counter to us at this point.

At this point. But they were all-in on, “You use apps, that’s what they’re for”, this idea that you’re going to collaborate on the web — I’m not saying they hurt you, I’m just saying there’s a reason Figma only worked in Chrome for a long time, for example.

DF: Apple reasonably was concerned about battery and device performance, and took a very vertical approach as they do with everything, and also was patient — just like we’re seeing now with them. When it became the right time, they added in collaboration to many other surfaces and figured out how to make it work with the cloud but I think they showed the importance of design to the world in a way that had never been so vocal before, and it raised the level of the conversation. You could argue Microsoft at the same point was also really leaning into design, but they weren’t as vocal — they didn’t have Steve Jobs talking about “Design, design, design”, they had “Developers, developers, developers”, it’s just a different tune.

Work in Figma

Yeah, that’s interesting. Is there any context, looking back now, where Figma makes sense for one person? Or is it really a product that only makes sense if you view it in this context of collaboration?

DF: A ton of people that use Figma use it individually, and I think it’s critical that you build tools that work for someone individually, that they can then graduate into a collaborative stance and use with their team. But you have to get the single-player experience right and then let it evolve to multiplayer.

So when you started going to market, what was your selling point? The tool itself, the accessibility, or was collaboration the key from the get-go?

DF: When we first did our closed beta, multiplayer collaboration didn’t yet exist in the product. It did have sharing, and that was very powerful — you had this one space to view your designs with your team, and people were doing that in very team-oriented ways. But early on, things like our improvements on vectors, or the simplicity and quality of Figma, were more the differentiators — and then design systems with a unique component approach, and then multiplayer, and then many other things. We also got a lot of minimalists in our early user base — folks who believe in the cloud and believed in minimalism, because we didn’t have all the features.

It was interesting just to see that early base of users and how successful they were — two of our earliest customers were Coda and Notion — just kind of wild that those were two of the first customers we had. I don’t even think Shishir [Mehrotra] at Coda knew that at the time — I once brought him in to talk with the team about platform strategy stuff, and I mentioned this offhand as an intro comment, and he’s like, “I was what?”, so it was a fun group to be around.

How much do you think Figma has evolved with your customer base, as opposed to Figma actually influencing your customer base and how they evolve? Did your customer base naturally become collaborative and realize they needed Figma, or did Figma introduce them to working in a more collaborative manner that they hadn’t considered because the tools weren’t there?

DF: There was definitely a period of adaptation, some people got it right away, for others it was over time. Our first big marketing moment — I remember there was a site, Designer News, sadly I think it’s offline now, and there was a comment on the launch thread, “If this is the future of design, I’m changing careers”, or someone said, “A camel is a horse designed by a committee”.

But we went deep on anyone who had really positive or really negative sentiment around Figma — great, let’s learn from all of it and adapt as we need to, while also having our own points of view and pushing for them. Customers have always been inspiring to us, we’ve tried to take feedback from everywhere — support tickets, in-person conversations, formal research, sales, social media — for a while, social media was a great signal, it’s not as good a signal as it once was. Our user forums, everything, and data analytics.

As you get there, you form a picture or view of the world, you play anthropologist and understand what people truly need and sometimes the moment just changes. FigJam, for example, was a product we introduced right after the pandemic started, I’d always wanted to make a whiteboarding and diagramming product — I saw that use case in the wild, it was significant, I felt we could make a simpler tool. But rightfully, the team was skeptical, always going, “Is this the right time? We have a lot of other stuff to do to make Figma great”, that debate stopped with the pandemic, when our user base wrote in en masse and said, “Please, please give us this product”.

We need a whiteboard, yeah.

DF: Yeah. We started seeing that use case everywhere — people treating Figma like a shared space and the shared-space part of Figma is something we’re doubling down on.

Was that the real turning point, “This is where work is done”? I’ve called Figma the operating system of design, in that everything sits on top of it and below it, but it’s the common layer, does that resonate? Is that the moment that became much more real?

DF: It was happening already in many ways, we were doing it ourselves, seeing it with our customers, but the pandemic is when everyone started telling us, vocally, “Lean into this”.

There’s so much more that’s possible now as we bring more mediums to the Canvas, more expression to the Canvas, and let people truly get what’s in their heads onto one shared Canvas — to collaborate, but also riff, see a bird’s-eye view, and directly manipulate. AI is great, prompting is great, you should be able to do it in Figma — and you can now, with our agent, but you can’t filter all of creation through the lens of AI. If you have an idea, or many ideas in your head, you need to get them out directly too and also you have to iterate to get to an exploratory place.

Too much emphasis right now is put on “I’m working with the AI, the AI wants to go a certain direction, and I’m going along with it”, it’s almost like, “Is the AI using you, or are you using the AI?” — sometimes it’s unclear. AI is a tool people can direct and work with, it can resolve tedium, but you also have to push, you have to be the out-of-distribution force, because AI is trained on the distribution, and the most interesting, differentiated work will be out of distribution by definition.

The Adobe Acquisition That Wasn’t

So I have questions about that, I have questions about AI, and questions about Canvas, which is a big focus of what you’re talking about at Config this week. But I want to do a quick side tour, because I must, another very famous single-player design company, as I mentioned, is Adobe. The Adobe acquisition was announced in September 2022. I’d written — we don’t have to spend too much time on this, obviously it didn’t happen, so in some respects it’s not that important — but by that point—

DF: Yeah, but it felt like it didn’t happen for a long time, those 16 months felt like an eternity.

That’s right, which I do want to ask you about, get your point of view on. But one thing I’m curious about, I actually remember where I was when this happened, I’d written several times at that point about generative AI, particularly images, the AI question loomed very large to me when that news came out. But that was still a few months before ChatGPT had launched, so this was more burbling under the surface.

To what extent was AI part of the Adobe conversation? There’s a very plausible story that it wasn’t part of the conversation at all — you were the operating system for design, the operating system can disintermediate all the products that sit on top of it, which from Adobe’s perspective was a strategic problem. They had a huge hole in this space, Sketch had already taken that whole space on the single-player level, so I thought it was an obvious acquisition for Adobe, aside from all the AI stuff, just looking backwards. Which interpretation is correct?

DF: Probably both. I think Adobe was super excited about AI and understood its potential and importance, we had plenty of conversation about that, but it was not, I think, the impetus or driving factor for me though in making the call of, “Do we sell or not?”. I had no idea, would AI would 1/10th, or 10x, or 100x our business? I was in my head trying to play it all out, and as we’ve seen, it’s hard to play these things out. You kind of know what’s coming, but knowing when it’s coming, and the second-, third-, and fourth-order effects — that’s hard. And this is pre-ChatGPT, so imagine trying to play out the next five, six, seven years from that point, that made me much more receptive to a conversation.

That makes total sense. For Adobe, I don’t think it was the controlling factor — again, you just made tons of strategic sense for them. But for you, it’s like, “$20 billion is very certain and everything else is very uncertain”, that makes a lot of sense.

DF: Another contributing factor was that I was excited about the opportunity to think about Adobe’s Creative Suite from first principles, and go back to the user’s problems.

Yeah — it’s missing the layer that Figma provides, the thing that actually ties it all together.

DF: There’s so much expectation from users of any software that’s been around a long time. There’s a need that reinforces itself to “Add, add, add”, versus thinking, “Okay, we’ve learned a lot — how do we reinvent from the start and think about things in a new paradigm?”. Looking back now, AI is clearly going to be — and already is — a tailwind for our business, it’s TAM-expansive in huge ways I probably never anticipated at the time, it’s also interesting from the Adobe frame, because I’d challenge the way you framed it earlier.

Please do.

DF: Adobe acquired Macromedia, and through that got Fireworks — and Fireworks was really the predecessor to Figma and Sketch, but not a focus for Adobe. They had different Labs projects, but this was not their core, their core was creativity — for Figma, our core has always been design, those were different when the Adobe conversations were happening.

Art vs. Design

Explain that, because I think I see what you’re saying, but people would usually conflate them — creativity and design.

DF: The even bigger question, for the philosophers and art-theory folks, is, “What’s design?”, “What’s art?”, how do you differentiate design versus art? It’s muddy, but design has an aspect of problem-solving, it also has creativity. Art, I think, is a lot of things — you can get endless definitions of design and art — but I think of it as trying to take an emotion, idea, or concept and communicate it to someone in a way that really affects them. That’s not best framed as problem-solving, whereas design is.

How about this definition: art is an expression that it’s meant to be consumed by the end user, and design is meant to serve the end user.

DF: Well, I don’t even know if you should define art as being for an end user.

Yeah, good point.

DF: For me, one of the definitions I lean on is that design is where problem-solving meets creativity. Figma has always had people using the platform for creative use cases. But now you fast-forward to 2026, and design, creativity, media, in some ways art and in some ways not, and advertising — it’s all kind of merging together, it’s all one thing in a way I wouldn’t even have said in 2025. If you believe we’re in an attention economy — you experience this every day — and you believe you have to have a differentiated voice and really have a point of view in your work to stand out, and you think the way people judge software is the design, that’s the differentiator, but you also have to grab someone’s attention, design and brand are so connected. It’s all really coming together in such an interesting way, because of these second-order effects of more creation happening in the first place.

A phrase you’ve mentioned, you said it earlier in this conversation, you’ve said it plenty of times elsewhere, is that AI draws from the middle of the distribution, and to be differentiated you need to be at the tails. That makes sense, but it’s funny because it conflicts with — go back to that user comment that’s deleted from the Internet, “Collaboration is the death of design”, do you see any tensions there? You talk about Adobe, creativity, tied to single-player, the genius of one person, versus, “We’re a group of people collaborating to get a design out the door”. How does that not end up in the middle of the distribution too?

DF: It’s more of a mindset thing for any design team are they trying to do the safe thing, are they tryigng to go for the least common denominator where everyone agrees it’s a good idea? Or are they trying to be daring and bold and take risk? What we’re going to see over the coming years is the market rewarding the risk-takers. And I wouldn’t say it’s enough to be at the tail of the distribution — I think you have to be out of distribution.

Is that possible? Aren’t you on the very edges of the tail?

DF: Yeah.

Fair enough.

DF: I think every email I get from your mailing list is out of distribution.

Well, thank you. I appreciate it.

DF: If you can get one of the AI systems to replicate your judgment and framework-building, I would love to see it.

I would both love to see it and hate to see it, so I guess it cuts both ways.

DF: Sure, I might love to see it in terms of wanting to know how you did it.

AI Headwinds

Well, it’s interesting for you, obviously. You mentioned a few minutes ago that AI is a tailwind for your business, I think it’s safe to say the stock market by and large does not agree with that, yet you’re there producing incredible results — you had a great quarter last quarter, your biggest beat yet. Do you feel you’re in the middle of trying to prove a negative here? What are the drivers of your business? Do you have some sympathy for the people in the market who are skeptical of you, or do they just not get it?

DF: Markets typically have a narrative they’re attached to, and the narrative can shift — and maybe it’s still not the nuanced narrative that matters, but this happens all the time. Markets are so impressive as a force, and I just don’t think it’s worthwhile to try to argue with a market narrative.

Are they normal distributions, and you’re trying to operate outside the distribution?

DF: (laughing) I like that frame. I just think that you show up, you do great work, you focus on the inputs, you educate to make sure people understand, and eventually that’s either appreciated or not, depending on how the narrative is going.

Right now the narrative is one of AI winners and AI losers, I don’t even think that’s nuanced enough, if I think more globally about software, there are many software companies and strategies that will work that are not necessarily companies and strategies that people would necessarily call AI winners today. I think about network effects.

Are you a network effects business?

DF: Collaboration definitely has properties similar to network effects, so in some ways, yes. And if you look at network effects not just in the social sense between people but also for marketplace liquidity — that is absolutely a network effect in itself, just to have liquidity in a marketplace, I would say that’s an AI winner.

If you look at the long tail of customers that are non-technical — I invest in companies occasionally, and one of them is Ambrook, an accounting-for-farmers company. I don’t think a lot of people in ag [agriculture] will be vibe-coding their taxes, they’ll care very much to have a human in the loop, for the certainty that this part of their business is going well and they don’t have to worry about it. I really believe Ambrook can provide a phenomenal solution there.

I also think liquidity of data matters — you need equity of data to create context, and context creates capability, if that’s self-reinforcing, you can get to a place where you have a virtuous flywheel that really helps in the age of AI.

Explain this in the context of Figma specifically, why does this provide a tailwind for you?

DF: I won’t go too deep, since it’s strategy, but the more activity people do in Figma, the more we can, with their permission, understand their needs and serve them better with capabilities. If we do that right, that’s a way to continually improve the experience for the customer and make it so they can do even better work, faster, in Figma.

How are you thinking about the models that undergird your various AI offerings?

DF: You always want to be in a place where models are swappable. We’re in an explosive, wild period of models constantly shipping, I went to bed last night and saw Sakana’s new release — I haven’t played with it yet, recording on Monday June 22nd just for reference. I didn’t expect that, coming out with their ultra model and their approach and just seeing the progress these labs are making, sometimes in a discontinuous way, is incredible. Right now we use a range of models and do some stuff first-party—

And these would be based on open-weights models?

DF: Some on open weights, some on very small things we’ve worked on. Overall, I think that there’s a big story around local inference that will happen in the future, as well as open weights and different models are good at different things, it’s incredible.

Is it fair to step back and say — from your perspective, which echoes a Microsoft perspective, or lots of other companies in a similar position — yes, models have to be swappable, customers don’t want to be locked in, but there’s also a self-interest position, you need to keep this data to understand customers better, and you need to not be giving that data to the models, who at the frontier need to not be swappable. Do you feel they have no choice but to come up into your space? Is there a perspective where Claude Design comes out and it’s like, “Yeah, of course that’s coming, because they have to own the consumer”?

DF: I think if you look at Anthropic right now — it echoes what we’ve seen from OpenAI over the past year, where there was a period when OpenAI was just building and releasing stuff in every area. And they, to their credit, have pivoted hard, made some hard calls, pulling back on Sora. That’s not an easy call after you do deals with major media players and have a huge launch and people are really enjoying the product, Sora was really cool, but going all in on code seems to be the right move for them right now, and it’s very respectable that they’re doing it. Anthropic’s going through a similar pattern, we’ll see what lasts and what ends up persisting.

That’s an interesting way to think about it. Did you feel pretty betrayed about the design thing — particularly when one of their executives was on your board?

DF: It’s complicated. Let’s put it that way.

Fair enough. I think it’s one of those things you could definitely see it coming.

Code on the Canvas

Tell me about Config. One of the products you’re going to announce is Code on the Canvas, tell me about that, and how it fits into the overall way you’re thinking about AI.

DF: Maybe to frame it up to start and dispel some of the stuff out there in terms of the way people talk about this — people on social media love to frame the “versus”, they’re always talking about code versus design, like they’re two different things.

To me, the work is not just vectors — it’s vectors, images, prototyping code, because you don’t always want to work in production, and production code, and production code needs to be across all your surfaces, web, desktop, all your mobile devices, new screen types, etc. All of that is relevant to your process, and all that process is design. So it’s super important to see it all as an “and” rather than a “versus”, I just want to make that clear because otherwise nothing else will make sense to folks.

If you think about it as an “and” and go all the way into what that means, then basically what you end up with is, “How do you bring these different mediums, these different materials, together in one place where it’s easy to go back and forth and get the benefits of each?”. For design representations like vectors and images, I think there are many ways those are very helpful — especially vector-based formats, for direct manipulation and precise control, in ways that code, which is structured, is not as easy to manipulate and mold. But code is also incredible, it’s got expressivity, full fidelity, it acts the way it will in production — hopefully, a prototype might differ from production — and you can have state and logic but you’ve really got to bring these things together.

So what we’re doing, based on the work we’ve done on Make, either from Make or by creating on the canvas yourself with code — essentially a code layer. You can have Code on the Canvas that pulls in from design if you want, and go right back to design — make changes and reconcile them back to code. We’re trying to make that all work seamlessly together, so you have a breadth of exploration while also having the collaborative aspects of the canvas and that bird’s-eye view.

Is one way to think about this that the question is that you can you eat development before development tools eat you?

DF: I think less that way, because my conceptualization of the moment we’re in is one that people are so eager to try so many different tools and materials — in some cases we’re going to be the best place to use those materials, in Figma, in other cases you’ll want to go elsewhere — and you might even want to come back to Figma afterward.

I’ve been thinking about this, the vibe-coding stuff is amazing, particularly in its ability to build scaffolding and get the functionality of an app and the user experience these tools build is hilariously horrible — it’s so bad, you really have to put much more of a heavy hand on it. When you talk about a phrase you’ve been saying regularly — that when execution is cheap, design and creativity are the edge, that’s very resonant to me in that actually conveying properly to the AI what you want is still a difficult challenge without it over-interpreting and over-assuming and spitting out a UI that makes no sense, and the design’s not just wrong at a pixel level, it’s wrong at a conceptual level.

I guess the question I have, and what I think you’re getting at with Code in the Canvas, correct me if I’m wrong — is that you guys owned the handoff between designers and developers where Figma was the common level where you could communicate back and forth, what’s happening, how it’s working. To some extent, if the developers are doomed, God bless them, designers rule the world — but did you accidentally erase your whole point of differentiation, which is owning that handoff between those two pieces? I don’t know if that makes sense, but it’s an angle I’ve been thinking about here.

DF: I don’t think developers are doomed, and I do think designers will rule the world.

(laughing) Both can be true!

DF: But I need to go all the way back for a second, when we started Figma, the first five years or so in market, a big part of our story, but also the ecosystem around us, was prototyping. And prototyping was not always with code, some companies tried that approach, but it didn’t really work at the time, because despite all the debate of, “Should designers code?” — debates that happen every year or two on Design Twitter, we would constantly see that designers did not all want to learn or take the time to code.

Now we’re in a world where it’s easier for designers to put their ideas into code. If you look at the prototyping aspect alone, in the Canvas, whether you’re working with production materials or prototyping, you need to be able to riff and explore and try things, and design representations are just one part of that, so is code. We’re also doing more launches at Config that add to that story. Motion, for example.

Yep, huge focus on this. You bought Weavy now you’re calling it Weave.

DF: Weavy, and now Weave, yeah. I love talking about Weave, it’s so cool. But Motion is actually coming from a hybrid of Figmates and a team we acquired called Modyfi. It’s something folks have always wanted — a timeline they can use in the Canvas and of course the challenge is how to do that in a way that doesn’t get in your way if you’re not trying to do Motion work. I think we’ve done a great job balancing those tradeoffs while providing a really powerful motion tool that’s much more intuitive than other approaches of the past and it’ll allow people go far more into expression, because it’s very hard to prompt and say, “I want the curve of the animation to be exactly like this”, the work we’re seeing folks do, even internally, with this motion tool is so incredible — I’m just totally wowed.

We’re also going hard on shaders, going all the way back to the WebGL conversation. It’s ironic, we were built with shaders all this time, but we didn’t give people using Figma the power to express in shaders. Now you can add shader fills and effects, and that unlocks a parametric option space to really explore this whole universe of effects, images, fills, and properties — and that’s even before interactive shaders, which add a whole new dimension, that’ll come soon.

We’re excited to bring all these materials to the Canvas so people can fully express and explore. And yes, if we do it right, it’ll be something they can then push to production — whether that’s pulling from Figma via an MCP, or more in the future, connecting to your codebase. We’re doing that with Make local right now, but we have much more to prove out there.

Acquiring AI Natives

I’m curious about that, because how do you think about customer acquisition? Back in the day you’d imagine starting, “Oh, Figma, this tool I’ve heard about, I’m going to make a design, and now I’m going to find a developer to code it”, now people can just get started with a ChatGPT or a Claude, and then it’s like, “Oh, this is really hard to design UI elements”, how do I back into something? How do you make sure you’re there if people are starting with coding in a way they maybe didn’t previously?

DF: I see people starting everywhere — that includes Figma, but also all sorts of other tools and places, and I see them ending everywhere. I see them ending in Figma to do the final iteration, ending in LLMs or other services.

What I think is essential for us right now is providing enough value always that the path to a great product is through Figma. Yes, optimally you can do that entire path through Figma as well, that’s a standard we should hold ourselves to. But we’ll continue to see people use a range of tools for a while, because these models are so underexplored. If we were to pause all development on models, a total moratorium, I think you’ve got like five years of catch-up on the application layer before the capabilities are understood and expressed through software. Every time I use these models, I find new capabilities.

Even there, though, is still the key for Figma is that it’s still the place people can work together? And that’s something AI hasn’t really solved, it’s kind of a one-on-one experience, but you need to figure out how groups can get jobs done.

DF: One area is groups working together to converge, I think groups coming together to diverge is also really important. Teams being able to work in all sorts of ways in the future is critical and also what are the things you’re always going to want as a team that are fixed, and what are your degrees of freedom? There’s so much we can lean into on collaboration in ways we’ve never been able to before, and make that single-player experience even better — because if we land all that together, you’ve got the collaborative layer, but also Figma is the place where you can just make anything you want.

That sort of leads to my question, which is, is the real Figma danger not that AI becomes multiplayer, but that individuals with AI disrupt multiplayer companies? And that’s why you still have to be relevant to the individual as well.

DF: I think it’s kind of a dark future if that happens, it’s one where folks are probably feeling pretty lonely — it’s also one where the tunnel vision you have when you’re building with AI is really becoming a problem for teams, I’m hearing this from design leaders everywhere.

There are different phases of AI adoption at these companies, the first phase is often, “We’ve got to use AI, let’s figure that out”, the second is like token-maxxing leaderboards — some extreme behaviors. The third, after they get people to adopt, is often “Okay, here’s your token budget”. In that second phase especially, where people go really wild with AI, it’s hard to get them to change their behavior after. A lot of people have this total tunnel vision of, “I’m building this one thing”, and they get really attached to it.

That’s the opposite of the breadth of what a great design process offers. If you’re going through the design process, it’s not that you should slow down necessarily, but you should go broader, and you should think. It’s essential that you actually think — not just wear a thinking cap, you need to be able work through yourself and have a mental model not only of the user and the experience you’re creating, but also cultural impacts, the broader system you exist in, what the user is expecting, all sorts of things. Going fast in the wrong direction is not progress, it’s a dead end, and it’s even worse if you’re collaborating, trying to bring five designers together and each one is viscerally attached to their one direction — now you’ve got design gridlock and you’re talking past each other.

So it’s imperative that we move away from this tunnel vision and toward the openness the Canvas represents. Maybe there are other ways too, but we’ve got to get away from tunnel vision.

AI and Path Dependency

On a personal level, how much do you feel constrained by the path dependency of having already built Figma? If you started out tinkering with tech as a kid, or even with the WebGL stuff, you ended up with a company. Do you ever have a part of you that’s like, “I’d just like to tinker with this tech again and not worry about whether it’s an existential crisis for this huge company I built”?

DF: I’m constantly tinkering. It’s my antidote to the non-verifiability of design — because there are verifiable domains and non-verifiable domains. Design is taste, culture, aesthetic, it’s constantly shifting, user experience is something designers can argue about in design crit for as many hours as you give them.

Unverifiability is the moat — that’s a good metric. The more something’s been argued about on the Internet, the longer a future it probably has.

DF: (laughing) The more you’re oriented toward questions than answers, I think it’s a good sign — it’s going to be harder for models to achieve it in a way that’s high-craft. And as a builder of Figma, that’s where the complexity and the interesting parts lie. The word of the year — not just this year, but 2025 as well — is evals, evals, evals. But how do you write the right evals for non-verifiability?

Aren’t evals, in some respects, counter to taste?

DF: Depends on how you do them, and who’s writing them, there are ways. It’s hard for LLMs to do well on aesthetics and user experience, like you said, and being surrounded by non-verifiability — when I go home and I’m finally unwinding at 11 o’clock, about to go to sleep, I’m not reaching for Netflix, I’m reaching for some model, and I’m exploring verifiable tasks, actually. Because I want to push the models on the unverifiable side we talked about all day long, but what can we do where it’s really verifiable and they have spiking capabilities?

Like vibe-mathing, for example, which oddly creates empathy for our vibe-coders. Because I vibe-math, and as someone who never went as far as I wanted to in pure math and wasn’t as good as others, I don’t know all the concepts the LLMs might be spitting out at me, so I have to learn as fast as I can — which is not fast enough, because the LLM is going through all sorts of stuff. It’s a great tool for learning, and super fun for discovery. And looking at the internals of models, how they work, understanding what you can and can’t determine, is also extremely interesting.

It’s all applicable in weird ways to Figma — you never know how. Even early stuff I did around understanding how to get models to have a broader range of outputs, and prompting strategies, I don’t think there’s one definition of the word “jailbreak”, but the things that got the models to open up more, exploring that direction, has really led me to understand models better, which benefits Figma in weird ways.

It’s super interesting. We didn’t get too much into the aftermath of Adobe, or the IPO, that sort of thing — but you talk about unverifiability and uncertainty, and that’s been the Figma story often, through things outside your control. It’s been interesting to observe, it really is quite an adventure of a company in many respects, really a unicorn.

DF: It’s been a blast, continues to be, and with the world shifting quickly, you can see it as chaos, or as opportunity — or both.

Are you glad you’re independent, or do you kind of wish…

DF: Oh, at this moment I’m very glad to be independent, we need to operate at such a speed and be able to pivot so quickly to make sure we update our priors.

Like the opposite of how you started, right? You started out with a two-year slog to even get this working.

DF: Totally. It’s so important now to constantly adjust as an org and make sure our processes support that, there are tons of things to do to improve there. But when people come to Config — which will be, as of the time this is released, I think happened yesterday, time’s weird on podcasts — I’m so excited. It’s going to be 10,000 designers in one place, and I get to spend time with the community and show them the stuff we’ve been working on. I think they’re going to love it and there’s tons more we’re working on, so stay tuned.

Very good. Dylan Field, nice to talk to you.

DF: Thank you for having me.

---

This Daily Update Interview is also available as a podcast. To receive it in your podcast player, visit Stratechery.

The Daily Update is intended for a single recipient, but occasional forwarding is totally fine! If you would like to order multiple subscriptions for your team with a group discount (minimum 5), please contact me directly.

Thanks for being a supporter, and have a great day!