As software gets easier to build, design becomes more important. At the time it felt like a tiny market. I think that so far AI is very much in the tool category. We talk about lowering the floor and raising the ceiling. The more time has gone on, the more confident I am in the designer's role. I believe that's going to be one of the critical roles in building software in the future.
It's definitely an incredible time. Ever since we started Figma, the exponential curve has been that more software is being created. Now it seems to be going vertical. I think design has the opportunity to differentiate software in this age of AI. We'll see new workflows emerge. Even from the start, over 80% of our weekly active users were outside the US, and one-third were designers. I'm excited about how we can activate those users more.
I think so far AI is very much in the tool category. We talk about lowering the floor and raising the ceiling. My point of view is to get more breadth in things you can try, but it still requires depth. We're in an uncanny valley where prompt to app doesn't give a well-designed app. The language people use, like 'locked in' or 'vibe coding', reflects the feeling of rapid iteration. At Figma, one of our values is play, and we want to create that rapid ideation loop for everyone. The disconnect is that you can prototype quickly but need to get to the finish line. It's also true for code. Founders are caring more about design year over year.
Can you talk about how you're thinking about the evolution of the actual models? There are two regimes: image diffusion models and code gen. Do you think one will work or both in tandem, or multimodal models?
It goes back to what is design. One definition is art as it applies to problem solving. The reason models aren't good at design is that on the art side you have diffusion, on the problem solving side you have LLMs, and it's not clear how to marry them. Designers bring research, understanding of user needs, cultural moment, empathy. I don't see models that can bring all those components together. The more time has gone on, the more confident I am in the designer's role. People who call themselves developers now might call themselves designers in the future, thinking more holistically about product experience.
There's a sense that LLMs will become super intelligent, but it seems there's a lack of agency. A large percentage of what a designer does is using agency to create on a canvas. Agency and judgment. I'm curious if that transfers over.
My hope is that humans retain the agency bit. We were hanging out with AI researchers who said models are more like hyper intelligent toasters. The opportunity is that the human designer understands people's needs and creates UI that matches. We equip them with tools to have a rapid feedback loop.
I'm seeing pretty much everything, which is exciting. It feels like we're exploding the possibility space. Prompting alone is like the telnet days of AI. There's a lot more to explore. I'm excited about different design patterns like AR, VR. Custom-designed vertical software per user? I'm skeptical. Snapchat and Minecraft are examples where low bar is optional. You learn them by watching others. There's a tradeoff if you shift interfaces too much.
It was a process. Evan and I started talking in late 2011. We asked why now? Technologies shifting: drones and WebGL. Evan vetoed drones due to regulation and hardware. We pivoted to WebGL. Then considered gaming vs tools, chose tools. I took a gap semester at Flipboard. We explored meme generation, which was a low point. But Evan made amazing text rendering that we reused for Figma. We looked at photo editing but realized the best camera is on the phone. Fireworks being killed gave us the thesis: as software gets easier to build, design becomes more important. The market was small but growing fast. It took a year and a half to get people using it outside Figma.
Field, D. (2025, March 13). Figma's Dylan Field: Exploring the idea maze, vibe coding, and the power of “locking in” [Interview transcript]. Y Combinator. CEOInterviews.AI. https://ceointerviews.ai/interview/1006131/
MLA
Dylan Field. "Figma's Dylan Field: Exploring the idea maze, vibe coding, and the power of “locking in”." Y Combinator, 13 Mar. 2025. Transcript, CEOInterviews.AI, https://ceointerviews.ai/interview/1006131/.
BibTeX
@misc{field2025_1006131,
author = {Dylan Field},
title = {Figma's Dylan Field: Exploring the idea maze, vibe coding, and the power of “locking in”},
howpublished = {Interview transcript, Y Combinator. CEOInterviews.AI},
year = {2025},
month = {mar},
url = {https://ceointerviews.ai/interview/1006131/},
note = {Speaker-attributed transcript with timestamps}
}
