Dylan Field: Is AI Replacing Designers?, Design is the New Code & Democratizing Creativity
- Cheaper code shifts software differentiation toward design, branding, and a distinct point of view. Field argues that app supply is growing exponentially while people use the same number of apps, making attention harder to win. “Design is the new code” means competing through products people care about; he explicitly says Figma still needs engineers.
- Figma’s growth challenges a blanket “SaaS apocalypse” thesis. The host cites 48% year-over-year growth and acceleration, while Field says Claude Design has not affected the business. Field emphasizes professional workflows and knowing the target audience: “It’s easy to confuse model progress with product progress.”
- Field has reversed his prediction that AI would raise the creative ceiling. “I was wrong,” he says of his earlier Config argument: AI lowers production barriers, but “AI doesn’t raise the ceiling, but people do.” His distinction puts the premium on human intention, experimentation, and the courage to create something outside the consensus.
- The failed $20 billion Adobe sale made prolonged uncertainty a central operating problem. Field describes 16 months of regulatory review and employees’ relief when the outcome finally became clear. Figma offered several months’ salary to people who wanted to leave, with a six-month wait before reapplying; only a small percentage accepted.
- Figma wants people to shape AI output, while disclosure can signal care and trust. Field says Figma aims to let people alternate between generation and hands-on work, while Weave connects model outputs into workflows: “You need a way to shape it.” Internally, a “written with AI” disclosure addresses whether the author knows what the material contains.
- Field’s optimism about democratized creation coexists with unease about automated discovery. The host cites an OpenAI release of over 700 math papers and calls it “the greatest day of discovery in the history of mankind.” Field welcomes the possibilities while admitting mixed emotions: “I don’t think it’s wrong to mourn the loss of a little secret.”
1. The failed Adobe sale lifted uncertainty before it required a reset
The host frames Adobe’s proposed $20 billion acquisition as an extraordinary exit that disappeared. Field says he respected Adobe and thought Figma had reached a size where it could join without being “completely absorbed.” But 16 months of regulatory review forced employees to hold two possible futures in their heads. “Having uncertainty for so long is really intense.”
Field calls the antitrust arguments “weak,” saying regulators fought past battles rather than examining where markets were going. His counterexample is Roomba: he argues that blocking its purchase led to bankruptcy, while the broader market was focused on China. He says the blockage did not make sense from the United States’ perspective. His broader objection is that prolonged review can damage either party, even when intervention is supposed to protect competition.
When the deal ended, the immediate mood was relief at certainty, rather than relief about avoiding Adobe. Field then offered an explicit choice: employees exhausted by the process, or who had joined expecting Adobe, could leave with a few months’ salary. They would have to wait six months before reapplying and interviewing again. A fairly small percentage accepted; some made substantial career changes.
2. Model progress does not automatically replace professional workflows
Figma launched in August 2012. Field says design is its center, but people also use it for marketing, advertising, slides, whiteboards, charts, and other visual work. He corrects the host’s framing: Figma had worked with Anthropic as a partner, but Anthropic announced Claude Design separately. He says Figma has MCP and cooperates with many companies.
Field describes surprising communication and suggests that extremely rapid organizational growth can create internal confusion. Asked whether the launch affected Figma’s business, he answers no.
The host cites 48% year-over-year growth and acceleration, despite having subscribed to the “SaaS apocalypse” argument himself. The host’s proposed distinction is founder leadership: someone who understands the user experience may be better positioned to use AI to accelerate and reinvent the business.
Field’s distinction is between model capability and finished products. “It’s easy to confuse model progress with product progress.” Professional design software requires years of workflow decisions and feedback across different tasks. The host compares its feel to “an instrument”; getting the mechanics right matters. Field is “neither optimistic nor pessimistic” about Claude Design itself.
Cursor is Field’s counterexample to inevitable application displacement: frontier labs pursued coding with their own models before Cursor had one, yet customer obsession and strategy produced success. The host then qualifies the SaaS argument: not all SaaS companies are good, and not all are dead. Field agrees.
3. AI lowers production barriers while people raise the ceiling
Field explicitly revises what he told users at Config a few years earlier: AI would let people do bigger things sooner and raise the ceiling. “I was wrong.” His current position is categorical: “AI doesn’t raise the ceiling, but people do”—through focused effort, boldness, and attempts at something unprecedented.
His software argument connects expanding supply to differentiation. He says app availability is growing exponentially while people use the same number of apps. With code easier to produce, products need distinctive design, branding, marketing, and a point of view. The uncomfortable part is expressing something outside consensus. “Design is the new code” does not mean engineers disappear.
An AI-generated deck makes the trust problem concrete: Field may like its sender yet wonder whether she wrote it or even knows its contents. Figma asks employees using AI in writing to put “written with AI” at the top. He expects trust to bifurcate, with doing something by hand one way to gain trust.
The host offers Photoshop’s layers, filters, and Kai’s Power Tools as an analogy: people feared what the tools would mean for photography, but they enabled new creative work. The host adds that camera phones were a bigger change, altering how people approached photography, and says it is hard to say whether that change was better or worse.
4. Creative tools must let people shape the generated material
The host asks whether human creativity—connecting experienced dots—is fundamentally different from AI. Field points to different reward and loss functions: a person could learn from one angry response to being splashed with water, whereas reinforcement learning would require repetition. Architectures might change, he allows; neuroscience remains poorly understood.
Field also distinguishes design from creativity alone: it combines invention with solving a user’s problem across branding, culture, and business constraints. The host suggests a prior aesthetic vision lets people use AI output as material to react against; without one, the output follows the model’s distribution. Field says the model adapts to the user’s vision, then describes Figma’s aim: make code material on the canvas and let people switch between building and working with their hands.
The host imagines media whose creators establish narrative and parameters while the final experience renders dynamically, as in a game. Field points to Weave, a node-based system connecting model outputs to other inputs. Generation supplies material, but “you need a way to process that clay. You need a way to shape it.”
Field describes a recurring novelty cycle: striking model outputs dominate feeds, then within days people recognize the tropes and ask how to move beyond them. He sees rapid progress in formal areas such as mathematics and cybersecurity, but much more gradual progress in creativity and design. “Maybe that will change someday.”
5. Greater creative agency arrives with mixed feelings about discovery
Field’s example of inventive digital art is a webpage—a letter to Figma—that degraded whenever visitors left and returned, making attention part of the work. He also admires Holly Herndon making her voice usable “almost open source.” That raises a separate fascination: what belongs to an individual, what is collective, and what deserves ownership.
The final distinction is how people choose to use AI. Field sees some letting it think for them; he prefers using it to deepen understanding, learn, or attempt work previously too difficult or time-consuming. “My whole life has been about figuring out how to think.” Everyone can create, he argues, given willpower, curiosity, and willingness to work.
The host cites what they describe as the previous day’s OpenAI release of over 700 math papers on well-known problems, calling it “the greatest day of discovery in the history of mankind.” Field feels excitement and sadness together: progress opens possibilities while diminishing the mystery of unsolved problems. “I don’t think it’s wrong to mourn the loss of a little secret.” The full shape of the coming changes remains unknown.
Full transcript
Dylan Field, welcome to the All-In interview.
Thank you. Thank you for inviting me.
We’re here in the beautiful city of Napa, Yountville, California.
Yeah, you’re already three drinks in.
We had a couple of drinks and decided to take some photos.
Yeah, why not? Why not?
Well, what are you doing here?
As your wealth grows, have you noticed how much more complicated things become? Between investments, tax strategy, and estate planning, managing it all is a full-time job you didn’t ask for. Creative Planning can help. Their coordinated team of investment managers, CPAs, and affiliated estate attorneys can do much of the hard work so you don’t have to. Creative Planning, where wealth works together. Go to creativeplanning.com/allin.
I was at this event that you’re also going to tomorrow.
I’ll be at the event tomorrow—the same event.
We mainly talked about design and product development. I connected a little to security; I thought that would be good. But in general, we talked a lot about how product development evolves.
From my perspective, you’d expect me to say that design is the new code. Coding models are getting better. I’m not saying you don’t need engineers. We definitely need engineers at Figma. We love our engineers. But the nature of programming is also changing, and with change comes increased value in differentiation at the top. It’s design, branding, marketing, courage, and the emotional reaction you’re able to evoke. I think a lot of people don’t notice that.
We also talked a lot about artificial intelligence and things like, “Okay, you did something on stage today. You spoke to Nikesh.” I’m just retelling the interview. Sorry, it’s wonderful. As if that’s good—as if we can fast-forward. But I want to go back.
You’re the CEO of Figma, a public company. Maybe you can briefly talk about Figma. When did you start the business?
We launched Figma back in August 2012. The short answer is that we want to make it possible for the best teams in the world to design, build, and develop the world’s best software.
We’re also used for all sorts of things: marketing, advertising, slides, whiteboards, charts—anything. If it’s visual at all, someone is probably doing it in Figma, even if it’s not intended for that.
People call it design software. Is that a fair characterization, or does it not cover the full spectrum of what the service is?
Design is truly our center and our foundation. When I’m in a product review and talking to the team, we talk about designers and what they want.
Design is also a very interesting term because there are about 100 definitions of it. I think a lot about the definition as creativity meeting problem-solving, though there are so many others you can find. Design can apply to everything from interior design to industrial design, product design, brand design, and more. People think about a lot of things when they think about design.
1. Adobe’s $20B deal blocked: inside the antitrust fight
Just to get acquainted with the history of the business: you started in 2012, raised venture capital, the business took off, and you sell enterprise software. You were doing really well, and then Adobe offered to buy the business for $20 billion?
Yeah, right.
Fast-forward quite quickly. What year was it?
That was a few years ago.
You’d been building a business for 10 or 11 years.
We always respected Adobe, and we still do. They’re an incredible generational company that started everything from printing to creating Photoshop as it is and the entire graphic design revolution.
By the way, I’ve been working in Adobe Photoshop since 1990. I was doing graphic design before layers. We didn’t talk about whether you were using it before layers, did we?
Yeah. I did graphic design and press releases as a kid on computers, as a service for people in the neighborhood, small businesses, and so on.
So cool. Wait, how did you get access to it?
I was in school, and then I had—
I was on a Macintosh. I got my first original Mac in 1984, and I’ve always been fascinated by graphic design. It was something I loved to do as a hobby. You and I have never really talked about this.
I was using Bryce 3D and all these early 3D tools, and then Kai’s Power Tools came out. Kai’s Power Tools was really cool. It reimagined Adobe Photoshop. For me, it was an incredible expansion of the creative potential of what could be done with Photoshop.
I’ve been a big fan of Photoshop since the very early days, and a big user as a personal hobby. I’ve made money doing graphic design and stuff.
That’s great, dude.
Photoshop certainly seems like a very mature product—maybe even more mature since so many different parts have been added to it over the years. That’s the nature of many software products: they get bloated over time if you’re not careful. But there are so many possibilities with it, and Adobe as a company has been very inventive. I’ve always respected them since childhood.
Absolutely. It seemed to me that this was a time when we could join in and not be completely absorbed. If we were smaller, I felt like maybe we would just scatter. We signed the agreement. We said yes. Then various governments around the world decided they didn’t like it.
The antitrust commission intervened and blocked the deal.
Yeah. Antitrust is kind of a broad term, but there’s the United States, the European Union, the CMA. A few years ago, it was a completely different atmosphere than it is today. Adobe would have to get rid of a major competitor. They said a lot.
But I think the most important thing for us was that we just kept pressing the gas. I thought everything would be as if we were just passing through, but after a month or two, it became clear that wasn’t the case. It became kind of normal. As an organization, we had to push ourselves, deliver, and create the next things. If we don’t push ourselves and keep our foot on the gas, we don’t know what’s going to happen on the other side. We have to make sure everything is okay, no matter what.
So, it’s 2023. How many employees did you have?
There were about 500 to 1,200. I’ll have to check again.
When I heard this story, I was captivated. It seemed like I saw you on March 24, 2024. I don’t know if you remember me mentioning this to you, but to me, the story was about telling 1,200 startup employees about one of the most incredible exits in history: this company’s $20 billion exit wasn’t going to happen.
What was it like when you stood up in front of everyone and said, “Guys, this is not going to happen”? How do you get everyone to refocus on building the business and continuing to move forward when it seems like the floor has fallen out from under some people’s feet?
I think you have to keep in mind the length of the regulatory review, which definitely needs to be fixed somehow. If you’re in a situation where it’s actually anticompetitive—and I still don’t believe the Adobe case was worth anything—many people on the internet would probably say yes, or at least they would have at the time. But I found their arguments weak. They fought past battles without looking to the future and what it would all become.
I’m not saying this because I’m not excited about independence. I’m excited to be running Figma. Adobe’s plan was to simply manage Figma. Anyway, running Figma makes me happy. Regardless, we move fast. I’m glad and grateful for that.
But back to your question: it was 16 months of regulatory oversight. Sixteen months of regulatory oversight is a punishment for any company. In a real anticompetitive situation, it could be very damaging to one of the parties, or both.
I think Roomba is a great example. Roomba should have been purchased. It was a small deal, not some crazy deal. They didn’t let it pass, and because of this, Roomba went bankrupt. The entire market at that time was actually focused on China, so from the United States’ perspective, it didn’t make sense.
Anyway, I digress, but I think this is an important point. There are political things here about ideals and idealism, and that doesn’t always practically help America.
It’s also important that almost all markets are now more expansive than they have been historically. Almost anything that looks like “Oh, this is a market” isn’t actually a market. The market is artificial intelligence.
Yes, I mean, that’s right. This is a huge market, and this deal represents 1% and 0.2% of the market. Do you define the market by looking back at history, which is the only logical way to do it as a regulator, or do you do it as a technologist, looking forward? It can be very different in how you think about markets and their size. Anyway, it all depends on the lens.
It was actually the first day of winter break, which is quite funny. We told everyone, “Okay, enjoy the holidays.” We actually have many fewer visitors during the holidays because we’re working on the product.
It was Friday, and over the weekend we realized, “Okay, this is going to end sometime.” On Monday morning, we sent everyone a message: “Hey, everyone, at 12:00, join a call if you want. Here’s what’s going on, and here are the full details.” First of all, so many people showed up.
It was great to see the turnout. It was the first day of vacation, actually. Secondly, I would say the mood was one of relief—not relief that there was no Adobe, but relief that there was no uncertainty. Having uncertainty for so long is really intense. People have to have 2 worlds in their heads; they’re not entirely sure what the future holds for them. The mood was just so much lighter.
And then, yes, I think your question is right: How do you make sure everyone agrees? We did it like this: I posted a video in the company’s Slack and sent everyone an email saying, “Hey, if you’re tired after working really hard, if you joined during this period thinking you were going to work at Adobe, and actually a startup isn’t for you, please know we’re going to work really hard. No shame—we’re still on good terms. If you want to leave, here are a few months’ salary, and you can just let us know. We’ll make it very easy.”
“You have 6 months before you can reapply. That’s the only condition. In 6 months, you can reapply and we’ll re-interview you.” It was a fairly small percentage of people who took us up on it. Overall, it was great, and some people ended up making complete career turns and going down wildly different life paths. I think it was great for everyone.
You continue to build the business, it grows, you go public. I want to quickly jump to the beginning of this year. You guys announced a partnership with Anthropic, Claude Design.
Not really. We worked with Anthropic as a partner, and Anthropic separately announced Claude Design.
Oh, sorry. Yes, that’s true, but you had a connection with them that you announced earlier this year.
Yes, we have MCP. We cooperate with many companies, but yes.
2. Is AI killing SaaS? Frontier Companies vs. Figma, the “SaaS apocalypse”
Then they announced Claude Design. Was it a surprise to you when they did that? How worried were you when they released Claude Design, compared to what you were proposing? The market looked at it and said, “Oh my God, Figma is dead. Claude is going to crush Figma.” The nature of online discourse right now is that we’re back, it’s over. It’s just a cycle, right? How many times have you been declared dead in the last month?
Yes. Maybe a bunch of times that I don’t even know about.
Yes, and everyone’s used to it by that point. When you’re declared dead so many times and then you’re actually not dead, you realize at some point that maybe it’s not true.
Social media isn’t really reality, but, yeah, I was generally surprised by how this kind of communication happened. Usually everyone wants to do good. There are some exceptions, but for the most part, that’s what I assume about people. They want to win; they want to do good.
Organizations scale very quickly. Anthropic is an example of an organization that has scaled that quickly, and that can be very confusing internally. Because of this, strange things can happen. I think we’re on the other side of some strange communication.
Has this affected your business?
No, no, not that it matters.
I read your quarterly report after that announcement. You guys grew 48% year over year, and the business actually accelerated its growth.
Yes.
I look at it, and this was when everyone was generally talking about the SaaS apocalypse. I was a big SaaS apocalypse subscriber. But I think there are many ways to differentiate SaaS companies. One of them, in my opinion, is being founder-driven. There’s a very important built-in user experience and an opportunity to build on that, especially if you know how to use AI to accelerate and reinvent your business as a founder. But your business has accelerated.
Yes. No, it continues to work well. I think the hardest thing about modeling companies and product is: What is the product? Is this a model? Is this the user interface on top? Is it some hybrid of both? If it’s a hybrid, how do you weigh it? Anyway, you’re talking about something like Anthropic and OpenAI.
Yes, I mean Gemini. For example, is it a coding loop, or is it a product like CoDesign that you just mentioned?
I think it’s easy to confuse model progress with product progress. The reality is that a tool like Figma took a very long time to create, and a lot of time to make sure we were setting up the right workflows for our users. We still have a lot of work to do. Our users will tell you that. We’re always trying to make it better.
I think it’s not surprising that it wasn’t an instant hit or anything, given that there’s a lot of work to do on the basic user experience and workflow to meet the needs of professionals. Now they’ve turned CoDesign into a broad product. They’re trying to do all the things you’ll see, like posters and graphic design stuff, when you go to their homepage. Maybe they do well in some areas and not so well in others. I’m neither optimistic nor pessimistic about the product. I think it’s just hard to create good design tools for professionals.
There are many things to design, and many design categories. The feeling is like an instrument. You have to get it right. You have to get the mechanics right.
Yes. We’ve been emphasizing every detail of that and getting a lot of feedback over the years about what people like and what they don’t. It depends on where they come from, their task, and what they’re generally trying to do. There are so many scenarios.
Overall, I think there are more modes for software than the SaaS apocalypse suggests. It’s not that software as a category is binary, good or bad. There are many ways to sustain yourself and have a growing business in the long term as a software company or an application-layer company. Part of it is specializing as much as possible in your target audience: know your audience and create for them. That’s what matters.
A great example of this is Cursor. Cursor definitely shouldn’t have evolved the way it did. You have all these frontier labs chasing them with their own models, before Cursor had any model. But Cursor is truly obsessed with the user. They saw their advantages, developed a strategy, and made sure they were doing the right thing for the customer. They achieved incredible success. I think that’s amazing, and I think it’s an example that will be repeated for many different types of companies. SaaS is not dead.
Well, I don’t think it’s true that all SaaS companies are good. And it’s also not true that all SaaS companies are dead.
Quite right.
I’m going all in. Conservation is often called anti-growth. The Nature Conservancy disagrees. They believe that protecting land, wildlife, and access to nature goes hand in hand with economic opportunity. For 40 years, they have moved at the speed of the market to protect lands at risk while supporting local communities. With 9 million acres protected in all 50 states, The Nature Conservancy is proving that it delivers long-term results that benefit communities and nature. Learn more at conservationfund.org.
3. Creativity in the AI era: why taste and boldness win
In the world of agent-based AI, I have this ability to give a hint, like a human, and say, “I want to create something that looks, feels, and does X.” This magical technology takes off and does it, and as a result, I get a product. Sometimes the product is great; sometimes the product is terrible. But these tools are increasingly able to do that.
Historically, I needed a product manager, a designer, and a bunch of engineers to deal with the actual creation of a product, whether it was a software product or a hardware product. It increasingly seems that AI is using more and more of these tools. Can you share your thoughts on the future of human creativity in this era, this future era we’re waltzing into—the singularity we’re entering? Where does human creativity play a role, and where else do we need to fill in the gaps to bring human creativity to fruition?
I think we’re lowering the bar with AI. We’re making the entire production process easier, whether it’s code, motion, video, design—if it’s a digital process, we’re lowering the bar for it to happen. Everyone can be a part of it, and it’s more democratized.
Ironically, even a few years ago, when I said this at our annual Config user conference, I believed AI would raise the ceiling. You would be able to do bigger things sooner. I was wrong. The ceiling remained unchanged. AI doesn’t raise the ceiling, but people do.
Creatives and designers raise the ceiling when they decide to focus, invest their time, be bold, and try to do something that no one else has ever done before. At this point, the question is: Do you care about being at the center of the distribution? Do you care about the average score? Is that enough for your task? Do you need something that shows you care about this thing, that you took the time to create it? Do you need to go beyond the minimum?
In the world of software and product design, we’re seeing exponentially more software than ever before. That’s exactly what one would expect. The expansion continues. You look at the App Store numbers, and they’re growing exponentially. We see that app usage isn’t changing. People are using the same number of apps, but the number of apps available is greater than ever and growing.
And I think it just goes to show that you need to differentiate if you're building any software. How do you do it? It’s not code, because now anyone can create code. That’s not the difficult part anymore. In my opinion, design is the new code. It’s the top layer. You have to make your design unique, and you have to have a point of view, good branding, and marketing. I think that perspective is what’s most missing. This is the human component.
It’s not just the human component; it’s also the part that people are uncomfortable with. I think people are afraid to say something that isn’t the consensus.
This is so true. That’s what the greatest artists, business leaders, and politicians have always demonstrated: the willingness to be bold, to say and do things that no one else would say or do. That’s what changes the world. I’d also add comedians. Comedians are underrated. Last year, we had Tony Hinchcliffe at our event because he’s one of the few guys willing to say and do things in this day and age of cancel culture. Comedians often tell the truth in a way people might not otherwise hear. They’re ready to say things.
It’s not about political correctness; it’s about whether you have a message that can rise above the noise and do so on a sustainable basis. Look at social media now: every unique point of view, anything interesting, spreads instantly. If you don’t have something that lets you keep creating content and a message that’s truly evergreen because it’s outside the Overton window, you’re not going to break through and get attention. The attention economy is ruthless. Is this also human creativity?
If you don’t do something that’s never been done before, you’re not creating something; you’re imitating something. You copy something, you photocopy something. Doing something new is creating, and novelty requires doing something that’s never been done before, by definition. That’s why you’re operating outside the training distribution. It’s not just about working outside the mean distribution you start with when you’re simply prompting without any advanced framework around it. You’re actually doing something groundbreaking and unique, and I think that will be rewarded by the market. Everyone will see it.
I think it’s also rewarded by the market, through organizations’ attention. I’m already hearing from many people—for example, today I received another deck created by artificial intelligence.
Absolutely.
I don’t even know if the person who created it, who I like, knows whether she wrote it. Does she even know what’s in her deck? At Figma, we made it simple: if you use AI in your writing of any kind, just write “written with AI” at the top. It’s very simple and easy to understand. People joked about it internally, but people do it. It means we can all be confident that when we see something, we know it’s been cared for enough to make it worth our time to process.
I think we’re going to see some kind of bifurcation when it comes to trust. One way to gain trust is to do something with your own hands.
Another important point is that if you have a vision or a point of view in your head about how something should be designed or look aesthetically or empirically, I think you’re in a great place to start creating with AI. It gives you something to overcome, like the problem of a blank slate, a blank canvas. You can react to it and say, “Here are all the things I don’t like. I’m going to fix these things. I’m going to change them.” It’s very useful to be able to exchange ideas.
But if you don’t have a vision of your own and you trust AI to make it happen, you’re doing it through the lens of AI. It’s as if AI has its own distribution.
The model adapts to it.
Yes, so you won’t get exactly what you want.
Now, if you can do both—and that’s what we’re trying to do in Figma—make code the material on the canvas, make it so you can start building. You can build something if you want, but when you want to think and work with your hands, you can do that too.
Yes, quickly switch back and forth. I remember the early days of Photoshop. I was an avid user from the beginning. People panicked when layers came out, then when all the filters came out, and when Kai’s Power Tools came out. Everyone thought, “What’s the role of a photographer now? This is destroying the art of photography, the creativity of photography.” But it unleashed a new era of creativity, because new artists could use those filters, tools, and layering capabilities to redefine what was possible on the canvas. It was, in essence, a new set of tools.
That doesn’t mean the old creative tools were a perfect match for the new creative system and tools. But this led to an explosion of creativity and human potential, and we pushed the boundaries in every way. It seems we’re entering this era again. I strongly believe that, and I think it’s partly what I described regarding the market dynamics. To attract attention, engage, and make people care, you need to do something new. You have to be brave and try bold things.
Resistance to change is real. There’s a great thing you can find on Google to support your point of view: search the mailing list for information about when layers were introduced. People say, “This is the worst thing that ever happened in the world.” You can find it if you Google it; it’s a publicly available mailing list, and you can see all the reactions.
Absolutely.
But apparently that’s the future. There are still photographers in the world who do incredible work, but the bigger change, to be honest, has been the camera phone. That’s what led people to stop buying cameras specifically for photography, because the best camera is the one you have with you. We started to approach photography differently with Instagram and camera phones. For a lot of people, that changed things, and I can’t say whether it’s better or worse as a result. But changes have happened.
There’s one view of human creativity that says it’s essentially a connection of dots in the subconscious, based on what you’ve seen or experienced in your life and your sensory inputs. From that input, you connect the dots and extrapolate a perspective on things the world maybe hasn’t seen before. Do you agree with this view of creativity, and is it different from what AI actually does? Again, I want to return to the uniqueness of humans in the age of AI.
One thing that happens with people is that they’re always looking for novelty. Our reward function is different from AI’s, and so is our loss function. If I opened this bottle of water and splashed it on you, you’d get mad at me if I did it again. With AI, with reinforcement learning, you’d need to repeat it many times. You don’t do that. Robot, stop digging me.
Maybe that will change over time with new architectures, but in general, the brain is different from the modern systems we’ve created with AI. I think it’s unknown; we don’t know much about neuroscience in general. I’d love to get to a point where we all know more about neuroscience, but something is fundamentally different when it comes to creativity.
I also think design is different from creativity. It’s not just creativity; it’s also problem-solving, as I mentioned earlier. The combination of those 2 aspects is fascinating, because you have a user whose needs you’re trying to meet in a creative way that connects with them. You also need to connect to broader systems—marketing and branding, culture, business issues, and so on. Keeping all of that in your head at different levels of abstraction, then trying to create a screen or a flow in a search space that can be a combinatorial explosion, is pretty cool.
This is quite fascinating. I’m personally very passionate about rethinking existing art forms and media, where artificial intelligence can enable more dynamic generation. The creator takes it to a new level and sets a bunch of parameters, but doesn’t necessarily render the final product. The final product, like in a video game, is rendered on the fly, with the user in front of it. The creative team behind a video game creates the whole thing: they build a storyline and a narrative flow. The implementation is much more dynamic, but I feel like it will find its way into many other forms of media.
We have products for that. In Figma Design, you can have generative tools and shaders where you can do a lot of that. We also have a product called Weave. Weave is essentially a node-based system. You can connect model outputs to other inputs, and so on. The whole idea is that generating with AI is fine, but you need a way to process that clay. You need a way to shape it. With Weave, people make the most amazing videos and photo works, and the results you can achieve with it are incredible.
If you have the intention, if you really want to do the work, you can always achieve a better result. New models appear, and people get this kind of AI psychosis for a few days. It’s like with algorithmic feeds: the best results rise to the top, and everyone says, “This model is everything.” Then, a few days later, it’s, “Okay, okay, I know the basic tropes of the model.”
And you adapt quickly. The novelty wears off. It’s no longer something that rewards you or triggers your dopamine system. Instead, you think, “Okay, it always does that. How do I overcome this?” This seems to be the continuous cycle we find ourselves in.
That’s quite exciting, because models are getting so good in these formal areas—mathematics and cybersecurity. IQ scores just keep piling up. For creativity and design, though, progress is very gradual. Maybe that will change someday, but I think the fact that these are untested areas has something to do with it. Maybe that’s the human role in the AI toolbox: to conceive, create, and design, and let AI do more of the work, while giving you the ability to do much more.
Is there any art or media that you’re particularly interested in that’s emerging right now?
I find it fascinating when people do new experiments with AI, especially in media. A designer-artist recently created a webpage that was like a letter to Figma. That caught my attention because I was seeing the Figma thing in my algorithmic feed. Every time you walked away from it and came back, it had degraded a little. That was, in fact, commentary on attention, and I thought it was really cool.
Oh, cool.
It was a work of art.
I thought so.
It was a webpage.
Yes.
The technique and the way they did it were really neat. It’s not something many others have done. I’ve seen a few other things like this, but overall, I thought it was a really cool way to do it.
That was also a very well-written letter.
I liked the letter too. I see things like that, and I remember people making AI art with NFTs back when NFTs were hot, around 2019 or so. The work was sometimes truly remarkable. Holly Herndon took her voice and made it so people could use it almost open source. That was pretty incredible and ahead of its time.
I think it’s a fascinating game between what is me, what is collective, and what is unique and worthy of ownership versus what is not. I’ve always been fascinated by this intersection.
4. Human agency, AI’s promise, and what comes next
Yes, that’s cool. A lot of the negative comments about AI are that it deprives humans of freedom of action and limits human potential. Could you talk a little bit about your thoughts on every human being’s potential to be creative, to be a designer, and how AI can perhaps augment every human being—not just “creators” in some elite class of artists in a special building, but every human being’s ability to create, design, and produce things in the world? Where are you at with that right now?
I think it’s even broader. As a human being, before AI and after AI, you have a choice. Do you want to create as a consumer? Do you want to take the easy way out, or do you want to climb a mountain, for example? What I see a lot now is people letting AI think for them. Less on Figma, but more in my interactions with the world—I see a lot of that now.
That’s one path you can take. Another is to use AI to deepen your understanding of the world, to teach, and to do things you would never have dared to do before because they would take so long or be so hard. You can really expand the potential ways you can live as a human being.
For me, this last one is the most interesting.
Yes. Why should I want someone else to think for me? My whole life has been about figuring out how to think. I just hope people will use AI to create good and be a force for good, rather than just joining the collective mind.
Everyone can be a creator. Everyone can do both. The main thing is willpower and curiosity. If you’re interested and willing to do the work, I believe in your truth.
Yes, that’s right.
And then I think you can do incredible things. We’re probably going to see a very different situation in the future, where things we can’t imagine will become possible, and things we take for granted will no longer be things people do every day. What exactly that will look like, I don’t think any of us can say.
But this is a moment when we’re both on the verge of major changes, and nothing struck me more than yesterday’s drop of math papers from OpenAI, with over 700 papers on all sorts of very well-known problems. It was a demonstration of something. I had a lot of emotions. I don’t know how to process them. I thought it was the greatest day of discovery in the history of mankind.
Oh, I mean, actually, yes. Something about it was brilliant and exciting, and something about it was a little sad.
Yes. I don’t know how to express my sadness, but it felt like we lost something and gained something at the same time.
I know. I raise my head, look at the next frontier, and think that these frontiers are brighter, wider, and more accessible than ever. So I’m very excited.
Oh, I’m excited too. I think the world is going to look amazing, and I’m an optimist. There will be many challenges, and we’ll figure them out.
Yes. Exactly how, and what these challenges will be in full, remains to be seen. But mathematics has a quality of mystery, like an unsolved problem. Is this hypothesis true or false? Why? Uncovering this secret is something special. I’m simultaneously excited about progress and everything it will make possible. I also don’t think it’s wrong to mourn the loss of a little secret.
Right. Right.
Dylan, that was a great conversation. Maybe we can have another drink?
Sounds good. Let’s do this. Thanks, dude.
Thank you. I’m going full steam ahead. I’m going full steam ahead.