Why Design Is the New Code | Dylan Field

A year from now. The same thing I think is very obvious now, which is that design is the new code.

If you believe that coding models will continue to get better, then you also believe that it's more outer-loop architecture — the decisions you make about how to go build a system that matter, not the code itself. The code is material, and the value rises to the top, and that becomes brand, it becomes product design, your marketing, your point of view on the world, and overall the way you'll shape products is through design.

Thank you for having me. Good to see you.

Well, most importantly and first of all, so we learned that this is already something we kind of knew, and then we viscerally learned it. We viscerally learned how I think designers, product managers — they're willing to deal with the terminal, the IDE. It is not the preferred way of working. It doesn't feel like home.

And they want home. They want a better way. And so I think that agent is a part of that. Code layers, which we announced at Config as well, will also be a part of that, and can't wait to be able to share more on that. But we also, I think, learned how you really need to just go back to the basics when building non-deterministic software, and this is also obvious. But building with non-determinism, building with models — none of it's rocket science. It's just a little bit different than working with determinism. And in order to do it right, you have to know the obvious things like, okay, you made a change. Did it get better? Did it get worse? Where did it get better, where did it get worse. And so things like that, we had to really dial in. We already thought we had dialed it in, but not enough. And based on all of our changes and improvements and the team's hard work, I'm really excited about this GA launch and being able to bring it to GA. It's faster, higher quality, works better with context, it just in general can understand better what you're trying to go for, and I think understands how to work with Figma better, which people obviously really want. And I think that we've also seen improvement on design system adherence, but more to go there and in all the places. This is still very much the start, and also it's something that we're really proud of.

Well, I think number one is design systems in general. I think that there's so much more to do there, and using the context of how you're designing locally in your canvas as well as what you've done in the past. We do some of that now, and we'll do a lot more in the future. And I think that, you know, these tasks are hard. You don't want to hallucinate, and you also need to be able to use the right things in the right times in the right ways. And the bar — this is one of those things where the bar for the machine is even higher than the bar for the human. You know, humans don't get this right all the time either, but we expect more from machines, and we should expect more. So we are really working towards that, like, you know, superhuman design system usage. But I also overall just want to point out, like, the entire goal is empowering designers to the max. Like, this is not a designer replacement. And I think there's some things where you can do tasks that are tedious and, you know, will take hours for a designer, and you can do them just like instantly now. And there's other things where designers with lateral thinking and, like, context and past experience — they just have these skills that models do not have, and are able to make associations with real-world sort of inputs that are moments in culture or the feeling that you're trying to evoke. And these are just things that I think we will continue to see that brand and product design and point of view are differentiators, and craft is a differentiator, and I think that the well-crafted experiences will stand out. And so agent is about empowering, helping you — it's not replacement. And working together with the agent, with multiple agents, multiple people, I think is the future.

Well, in the past I felt that — in the years past I said AI will lower the floor and raise the ceiling, and I had to kind of correct myself because I'm like, actually, that's not what we're seeing yet. We're seeing it lower the floor. It's more democratized to do something. But the ceiling doesn't seem to have changed. And despite, like, all the amazing math results and the formal domain results where you're able to formally verify something — you know, anything you can verify, AI is great at math being number one, maybe cybersecurity number two — you can test to see, you know, like binary yes or no, does this work. You get five designers in a room, they can't agree all the time on exactly what they should do, and that's what's interesting about design — there is not one solution, there's many solutions. And I think that overall, that said, we're getting a lot of center distribution. We're getting a lot of sort of average, the mean work being created. Another word for it might be slop. You know, it's funny though, because a lot of the stuff that we call slop, you know, it wouldn't be stuff we'd call slop a month ago or two months ago. And so as the models change, our definition of slop changes. Overall, I think that the craft that you imbue into something that you're creating — it is reflective of your thinking. It's reflective of your care, and because of that, people pick up on it whether they can articulate it or not. They can tell, and they will judge appropriately. So I do think it creates a lower bound that you have to exceed, but it doesn't define the upper bound. You define the upper bound. You define the ceiling.

I think we're in a period of time, and it's interesting because people, I think, are going through the sort of AI loop of like, oh my gosh, the models — they can do what? And then kind of coming out the other side, and sometimes it's repeated, but people are going through it at different intervals. And in a group setting, I've seen some companies that, like, have completed this loop end of last year, and some companies are just starting to go through the loop and, oh my gosh, what can I do? And so I think as people come and come out the other side, there is just this return to, okay, we still have employees. Like, people still have jobs. We believe that the folks that are at our company — I mean, I certainly do for Figma — they're great thinkers. And because they're great thinkers, I want them at Figma. And I want to hear their thoughts, and I want to challenge their thoughts. I want to hear their second, third-order thinking. I want to see how they're going to plan around corners and, you know, what their predictions for the future are and what their judgment is. And I want to push further with them and explore things that have never been explored, like you said. So I think it's back to that. And that's not to say that you can't use AI to help you there. I think that the danger is if AI is doing the thinking for you and you just kind of throw it over a wall. Whereas it's way better if you are using AI to learn faster, try things you've never been able to try before, and just generally be more daring.