Figma CEO on How to Get Good at Design in the AI Era | Dylan Field

If you're a PM and you think that your job is to make documents and slide decks for upwards review and alignment, you're going to love this new world that we're in. You get to make things too. People need to see that leaders in their organization are making things too. And that is what inspires and actually creates the inflection. We're entering a world where design is the new code. You'll be designing in a visual first way and you'll be able to do a pull request right to production. Direct manipulation on the canvas is just clearly superior to prompting and probably design direct manipulation is superior to code editing. As we move ahead, the world gets a lot more visually interesting, too. I'm hopeful that this will be like a renaissance period for that.

Thank you so much Peter. Appreciate you having me.

You can still do it. Not too late.

It's funny how Silicon Valley is so excited about taste. I think craft is also maybe overlapping but distinct as well, and then there's point of view. You can have taste as a person and you can build with craft and intention and still have a very clear point of view even though you were incredibly craftful. As you think about what are those different territories, taste is really about navigating the possibilities of what's out there and having preferences that are really clear that you can articulate, and also being able to help others understand what it is and what it is not that you're going for. Craft to me is really pushing past where others might push and thinking about things in their entirety and down to the micro and small level as well as the macro level. It's all the different levels of abstraction. You make sure they fit together. And then there's point of view. For point of view, you really want to be expressing through a product or a design something that is unique that you see in the world, and doing it in a way where you're almost bringing some insight or some take to life that is moving some conversation forward. Across many products, consumer, enterprise, software, hardware, the best products actually do have a point of view about the future. They have a point of view about how things should be and how people should live their lives or work. If everyone agrees to your point of view, you're probably not having much of a point of view.

Exactly. Yeah.

And I think it's fine to have some interplay there. In fact, it's really good to listen to users and customers to get to all sorts of local maximum, but point of view is maybe global maximum or it's going to get you to the next local maximum rather. Or the first one. Otherwise you're just kind of iterating around where you locally are. It's always interesting to ask designers about taste or ask them what good design is. First of all, the number of definitions you can get about design itself from a number of designers is amazing. Also, if you do a design crit with really experienced designers and you have five in a room and you argue through from an aesthetic standpoint or from a UX standpoint or both, what is the best design here on the table? You might get a hundred different answers from those five designers. So there's just so many different places you can go, and then it's about how do you actually narrow down that possibility space and traverse that tree properly.

I think that over time, absolutely you can. But there's also always the cultural influences of any given moment that really matter. I find that the most interesting points of inflection are where you're on a frontier and you're pushing further than anyone else has pushed already, and then you're rapidly updating in real time. You get a few observations out, whether it be on design or aesthetic or strategy, and then you make a leap rather than try to iterate one step locally, if that makes any sense.

Yeah. The funny thing about OpenClaw is you don't know who is posting about it, especially if you don't know the person who is actually being an agent or reflecting on what the agent actually did.

I think that models have already gotten to a place where their visual output, for example with the Gemini 3.0 release, 3.1, but even 3.0, if you prompted it the right way, complex prompt for sure, but if you gave it references, it understood, and you push it hard enough, you could get to some pretty incredible outputs. Those outputs are lighting up this possibility space, creating different places that you could explore. From there, though, I never found myself feeling like I got my answer one shot done. Even more than an average output or what you're referring to as purple slop, I actually had more opinions about the thing with point of view and direction to it than I did the purple slop. For average output, you're just like, cool, it does the thing. There's a lot more I would push it to do myself or a lot more expression I could have visually or experientially. But it works. Once you go another step further and you have some intent behind it, that's the point at which you get to interesting places where that dynamic loop starts to occur and you really want to push it further yourself. Prompting is not always the way to do it. A lot of times what you need to do is push it forward as a human in order to iterate to the right place.

I think it totally depends. There are times that we start with deep discussion. There are times where one of us is ideating away with pencil and paper and a notebook. There are times where we're in design or in a collaborative brainstorm, or we're literally in code or making designs. There are so many different ways you can start now, including from an existing setup or website or app that you already have. Wherever you start, our job is to make sure that as you hop around between the different stages, which were stages in a linear process before, between ideation, alignment, design, and actual production, you could start anywhere but you might want to go everywhere in that process. I want to enable that fully on the Figma platform because that's what I see happening more and more. It's not even right to call it a loop because it's not like you're always looping back to the same thing, but it's all these hops that could occur as you're really trying to explore. That includes a lot of divergence and convergence. The diamond shapes stack on top of each other and they could be from any point to any other point in terms of where the artifact is or what it is.

That's the beauty of digital, you never have to say it's done. People trained in industrial design or the physical world try to go through a process that can complete, and at the end you're shipping it off and if you don't like it, too bad because you cannot update it. In the software world of bits instead of atoms, we get to update constantly. It's an amazing world to live in, what a luxury. But also, the velocity or perception of velocity is increasing.

You need to run, but you need to run towards the right thing. Keep the velocity but have some sense of cardinality, know where you're going.

Yeah.

You might start in design, you might start in code or somewhere else. If you're in code as a place to start making a prototype just to get something built that you can play with, cool. You can do that in so many ways. We want to make it so that you're then able to go at some point and diverge and use Figma for the stuff that is better to use Figma with, for example anything from spacing to color to properties of the canvas that you can directly manipulate. Being able to have that rapid feedback loop where you do direct manipulation on the canvas is just clearly superior to prompting. Code editing itself is superior to prompting, and probably design direct manipulation is superior to code editing when you're adjusting things like that. In terms of breadth, the canvas is a natural place. It's an infinite canvas where you can see all the possibilities of flows and different ways to iterate on a screen, all in the same spot. It's important to be able to go from design to code, from code to design. That round trip we have to make as high quality and as efficient as possible. It's really important to advance that forward and have that loop be really tight no matter where you start.