Dylan Field on Figma Make + Claude Sonnet 4.5: the era of the 10x designer is just beginning

purposes. Among many others, Suzuki and Toyota have had multiple generations of adopted son-in-laws to keep them in the family business. That's crazy. Wow. Well, let's bring this concept into enterprise software. Dylan is in the reream waiting room. Let's bring him into the TVP Ultra Dome.

His first virtual appearance, I believe, on the show. He's of course uh hit a gong live with us at the New York Stock Exchange which is uh which we a slight upgrade from Zoomly retired the gong but uh but thank you so much for joining us Dylan. How are you doing today? Thank you for having me back. Good to see you guys.

Great to see you. Great to see you. Uh what's new in your world? Well, I've I've been deep in it this weekend uh with Cloud Sonet 4. 5 if I'm going to be real with you. Yeah, in Figma make kind of still swimming in that. What? Well, yeah. Give me your initial reactions. Like, do you care about the benchmarks model card?

Are you tracking like the the meter, like how long a task can run? Are you more like just chat with it like it's a person, have really short back and forth exchanges to get the vibe or the texture of the of the words that are coming out?

or you put it on a really critical business task and then benchmark it against one of your top developers or something like how do you evaluate a new AI model as a CEO? Well, okay. So, there's the personal side and there's the Figma side and there's a CEO side maybe. Sure.

Um, so I'll start with personal, which I have not gotten into yet, but I love playing with these models, uh, testing them at their limits, seeing if I can, you know, get them into sort of weird headspaces. And, uh, that's just like a hobby of mine. What's a weird You're trying to oneshot them.

No, I I mean, look, it doesn't have to be a one shot or like single prompt. You know, the longer you go, the more interesting it gets. Yeah, you keep going in the models.

I'm seeing what's the weirdest uh what what's what's an example of like a you know, you don't have to name the model, but like what what kind of weird what's the weirdest headsp space you've got in a model in is they're just hallucinating, you know, more and more intensely type of thing.

Um, look, I mean, models are trained on the internet and the internet uh has some weird stuff on it and you know what do you find it in there? It's kind of fascinating and uh I won't give examples or definitely not companies or model providers.

Uh what I do instead is I try to like if I find a good repro case, I send it over to the models and sure that way they can figure out what to do with it.

Uh yeah, it's funny if you think the models are the models trained on the internet that means like the models effectively trained on stumble upon you know like that that level of range right which is quite quite wide. Uh there's a lot to the internet deep for sure.

So take the other specifically at the not even at the I haven't even gone there with 4. 5 yet another time. Uh but I'd say that um for Figma for Figma uh that is where my attention's been focused so far.

And with Figma make in particular, we've been just trying to figure out, okay, how do we make it so that um we're able to quickly pull in Sonnet 4. 5 to make and then evaluate, you know, are we ready for it? Do we have to do anything differently? Is uh this a good addition, a good improvement?

And um the more we play with it this weekend, the more we used it, it was so impressive how it made make better. Um everything from planning and just um thinking through and evaluating, understanding the code base that it was working on, whether it was like a shorter prompt or a longunning session.

um to thinking through uh just sort of like the way it should evaluate a prompt and uh giving better results and also you know we recently introduced this new feature where you can copy from make output into Figma design and we've since launch had a way to basically go from Figma design into make and basically copy paste in and that way you can bring your design uh into whatever you're coding and creating.

And the consistency that we're now seeing with 4. 5 on that part of it, that round trip, that's the part that's most incredible.

And it's so uh just kind of like gives me the chills a bit because we literally launched this feature where we copied from make output to Figma design last week, not knowing when we launched it that, you know, 4. 5 was coming, it would make it so much better.

And here we are and we're like okay well that's a good surprise. Uh but I guess the general point is just it's really nice to be in a place where and I think this is an important thing for everyone to be thinking about in software as the models get better you need to get better.

So for us, we look at the models getting better at Figma. We're like, "Okay, great. Figma gets better. All of our users um are going to benefit from this. " And I think it's really important that you're always in that sort of headsp space of how to make sure that you're setting up your strategy that way.

How do you think about the Figma make user base, the community? Uh right now there are so many different subsets of AI users. There's folks who can tell you the difference from one paragraph. Oh, that came from 40. That came from CL sonnet 3. 5.

Uh, and then there are people who are like, I got to try that AI thing out soon. Uh, is is the Figma make community in a place where 4. 5 Sonic can just be like a strict upgrade that there's no resistance to or is it something where you have to think about model switching and and leaving some choice up to the user?

Are there cost considerations? Like how much work are you putting on the user or the community and where's the right pivot point for that? Yeah, I think there's to your question about um sort of the diversity of use cases and personas that might be applicable here.

There's so many different users on our platform uh and types of behavior that we see. You know about uh you know a third of our users are designers, twothirds are non-designers. Design's always at the core.

And I I honestly think that just from my reflections on this weekend, one thing I keep coming back to is like I think we're just entering this era of the 10X designer where the designer is going to be able to do so much more than ever before.

You know, designers have knowledge not only of aesthetic, UX, and craft, but also the value is just moving up the stack.

And the designer uh will be in this position of leverage where they're going to be able to do um a lot in terms of product building and creating products with craft uh that just create joy and thinking of the entire system and then pulling everybody else in and helping lead the charge.

And uh I think that it's just this time where good enough is no longer enough. Uh good enough is mediocre. You have to be great if you want to make it so that you can win. And uh I think that the more power that designers have, the more leverage they have, then the more likely they are to be able to help a company win.

Um, and so that's just kind of like how I see things going in general. Um, and on the implementation side, uh, I think it's like not as relevant.

You know, you just kind of have a variety of ways that people, um, want to interact or different cycles of AI adoption people are at and you need to meet them where they're at and you need to prog. And so it gets back to the basics there.

How do you think about um the the vibe or the taste of the designs that are coming out of these models? I keep coming back to the the midjourney example where it just feels like David Holes did something different from just take the average or like or like you know minimize the loss function.

It feels like there's his artistic vision in there in every generation no matter what your prompt is and then you can do a lot of different things on top of it. And so it's almost this like collaboration between the artist David Holes and the artist who's the prompter.

And I'm wondering if you think that like like how does the flavor of design that's coming out of 4.

5 plus Figma make like where does the human opinion get inserted at each level and how does that like how much of that is actually happening like can you tell the difference in the flavor of design that's coming out of a certain model?

I I think um we're not yet at like a Dolly 2 moment for Z or a majourney moment for Z. That's happened a relatively the same time. Yeah. And uh you look at Dolly one compared to Dolly 2, you know, it's uh it's easy to dismiss and um I think design generation will get better.

At the same time, I think that the law of averages concept that you kind of mentioned um is sort of the reality today and likely will be the reality for a long time. Yeah. And what is critical is iteration, the ability to push in a direction and getting to great.

And if you're able to do that and you're able to uh advance your craft, if you're able to have a sense of what the culture is and what the business problems are and what the engineering challenges might be, bring it all together with a great design, user experience, aesthetic.

Um it's it's like it's not just the style, it's the entire system you have to hold in your head and figure out. um along with brand marketing point of view. How is it all going to work work together? And if you can do that, that's like where I think everything unlocks. Yeah.

How what's your read on sentiment right now in the design community broadly around AI?

Because on on our side as a business that that is constantly working with creative people externally obviously we have people internally but we have like basically as much demand as ever for like talent talented creative people like we don't we have a ton of work that we want to do and uh but we also have an unlimited demand for Figma make and tools that help us instantiate ideas and there's a lot of stuff where we're like we just need to do the footer or the out or the buttons and like the faster we can get that out the better.

Yeah.

But but but there's this sense of um we have a insatiable demand for novel thinking and ideas and concepts and it's it's less a it's less a I can't tell if you're saying we humanity or we TVPN because I'm talking about both but but TVPN and it's less of oh is this possible like is making this animation uh going to be you know uh uh within budget it's more it's more like well what animation should we willing to pay willing to pay a lot for that kind of original thinking.

But I'm curious like where where you see the um kind of where where sentiment is at today because on the software development side it seems very clear as as when you can build software faster and cheaper. Well, we want more engineers to build more software because the world needs a lot of software.

And I think same thing for truly great novel creative work. Yeah. Um, well, first of all, I'm hearing a strong advertisement uh on TVPN for TVPN that you all are trying to hire. So, hopefully our viewers can like uh apply somewhere.

But the uh the thing that I would call out is I I think that um we're almost seeing a parallel to the AI talent wars with design right now. Yeah. At least amongst the companies that are really getting it.

Um the ones that have internalized the design is going to be how you win or lose and that's the craft, the details that matter. They are really getting aggressive. Not like meta AI aggressive, let me be clear, but very aggressive.

And um it's I think just again value moving up the stack and uh that's something that all companies will eventually figure out. I just hope that uh a bunch of companies don't figure out too late.

But yeah, I think Jeban's paradox as applied to talent that is adept at using models and uh doing amazing work and really pushing the boundaries of what's possible and what could be created and how much craft you can have, how much you can delight people.

Uh that's real and the demand for development is certainly not uh uh shrinking.

I think that um you know it's never been met and at Figma I mean we're doing headcount planning like we're hiring almost all areas as we go into this next year and in big ways and I think just generally you know our point of view from the start about AI at Figma um is like how do you both lower the floor make it so more people can come into the design process but also raise the ceiling make it so designers can do more and I think that's something that's a shared view maybe We were one of the first to articulate it if not the first but uh I believe that most of the products that are in the sort of AI space are trying to do the same thing.

They're trying to make it so that more folks are able to be part of uh this life cycle and also they're trying to make it so that people that are the experts are able to do more than ever before.

Do you have a view on like the shape of the design community as a function of maybe the amount of designers working in a particular organization? Are there are there more design firms than ever? Are there more solo design firms that are profitable and making money than ever?

Is the power law getting steeper or is the floor rising or both? Do you have any idea on like the shape of the design community and and the shape of how designers work together uh or just make money independently uh in the age of AI?

Yeah, I think um there's a lot of work here we've done on the sort of company side and we're seeing there just more and more people getting involved in design and more design hiring. Yeah, I think in that um more freelancer agency segment there's more research to do uh to understand is the same effect happening.

Um but I'll maybe think back historically um you know when there was this first wave of design hiring and people realized that design talent was critical to success. One of the first things they did was they started to snap up the agencies and they would literally just start acquiring agencies of great design talent.

And so uh that's how like for example Tannon Lax um became uh involved in then Facebook then reality labs uh and John Lax ended up leading up reality labs for a while and uh they got someone who was legendary design talent to be able to be at the forefront of their work on Oculus for for quite some time and um I I would not be surprised if similar things started happening Um but I also think that um you know back in the early days of Figma it was this question even of like how many designers are there in the United States in the world.

I mean the Bureau of Labor Statistics said that there were 250,000 designers. It's like well that seems wrong in 2012 but like how many are there actually? I don't know maybe it's like 500,000 not 250,000 but like is it like a venture scalable business? We don't know.

And uh but what we noticed was that values move up the stack and it was going to be the case that everyone was going to build up uh their design team and the design would be so important as this overall system changed. And I I think that that same thesis we held then in 2012 uh it has not changed.

It is the same thesis we hold today.

And I really believe that uh sort of this era of the 10x designers here, it's um just starting in some ways, but it's going to be quite incredible and it'll require more designers to step up as leaders and uh for them to lean into leadership roles that they may have otherwise not considered.

Uh but you know if they don't then I think it's going to be a dynamic where everyone's trying to get involved in design and they don't really know the guardrails or how to do it exactly and um you have a lot of people with a lot of thoughts but not many like railways built tracks built um for folks to go down.

So that's the challenge I think that many design uh oriented folks will will have and the need for design leadership this time. Shopify just acquired a design studio last month, by the way. Someone someone in our chat mentioned this. Um, what's your thinking around uh open source?

You obviously have a close relationship with with Mike over at Anthropic. He's on on the board and and obviously Claude uh has consistently led on on codegen, so they're a natural partner for Figma Make.

But is open source something you're exploring or or um just given that they're not uh there was for a while that was like commoditize your compliments but maybe the LM yeah I'm super interested to hear what you think.

Yeah, I mean, look, there's um I think uh first of all, I mean, Mikey and I from the start of just looking at uh him being on the board, we had to make sure that, you know, it was clear we as Figma have to choose whatever is best for Figma. Yeah. Um you know, as models get better, we get better.

But if we're not choosing the right models, then we're not getting better. So, yeah, on 4. 5, both of the eval choice. uh you know roll it out to users. There's gonna be a lot of other model improvements and of course open source is something we're always watching too.

And so uh it's exciting just how much is happening right now and I think that uh it's going to be just a fun next few months to see the continued model releases. Uh I'm I'm pretty psyched. What are your thoughts on hardware right now?

Um, that feels like a place where we're seeing uh a reemergence, a lot of hype around uh Meta's new devices, uh, potentially a new design surface as people design more um, more physical products uh, that are maybe enabled by AI.

It just feels like that could be kind of the next thing, some sort of like Cambrian explosion of hardware devices. Or maybe it's all just just AirPods forever or something. Who knows?

But I'd love to hear your take on like how is is is there is there a demand from you for new hardware devices like the the Waycom tablet is obviously legendary in the design community uh but it's a completely proumer professional device. Just what how are you thinking about hardware these days?

Well, I think in the Figma context, um, what we think about is, okay, what are the screens that we need to serve and our users need to design for? Uh, and that's why I was so excited to be able to check out um, Meta's new glasses and uh, also the neural band. They're the band is like underappreciated. So cool.

Yeah, people need to People won't it won't get a reaction until enough people have actually tried it because it's like it looks just like a whoop band. Like there's nothing that looks crazy about it until you actually try it and you're like, "Oh, you can fully see what's going on. " Yeah. Yeah.

And um but like let's be clear, I mean this was years in development uh many years. I mean they've been at it for so long now. Uh, and I think at the first time I got the demo was uh, almost a year and a half ago, year and a few months ago, and then I saw it again right before config.

Uh, but like the sort of new and improved version. Yeah. And it just takes a long time to go build these new form factors. Um, you know, the same thing is being true in auto. It's true, uh, you know, in wearables in general. uh it's probably going to become true in other surfaces that people will find.

And I I think that whether it's AR, VR, uh you know, curved glass, um you know, small screen sizes, large screen sizes, there's a lot of surfaces that designers now need to target. And we haven't even gotten to, you know, AI and agents and what do designers need to do to create context in these places, too.

I think in many ways, we're just in the MS DOS era of AI. Uh we are still figuring out how to explore this amazing uh laden space. You've got this LM. It's your spaceship, but you're in this like n dimensional space. And the best way to like figure out where you are, your compass is your prompt. Natural language.

It's like what interfaces will we built for that? How will we be able to figure out custom bespoke ways to do dimensionality reduction? Figure out how to help people navigate better as these models and the architectures fundamentally improve.

Um, and also, you know, what protocols between the models should exist and what's the best way to create those because I think that uh there's a reason MCP has really caught on so fast.

is extremely valuable and we just launched uh with Figma the uh Figma MCP for design uh as a remote service before it was only desktop and local and also we improved the capabilities and then we did it for make as well.

So we want to make sure that people know you're not trapped to make uh Figma make if you're going prompt to code and you're um making it so that you're getting to some great output and you're spending a lot of time on it awesome.

like you always had download code, but now you can just have your MCP uh connect to Figma make and pull all your code in. How is how is MCP adoption going? Uh it it it feels like uh there's a there's a world where like I was promised like the AI didn't need an API because it would just use the computer, right?

And I imagine that there's a maybe we're not quite there, but there's a world where my my AI just literally opens a browser, goes to think about. com, does whatever it needs to do just like a human. Uh, and yet MCP seems to be like what you're identifying with like the energy around it is immense.

Um, how are people actually using it? Like are there is there a killer app, killer feature, some story where you've been really excited about like the impact that it's had on an actual like design project or designer or firm or your company? I mean, you should definitely try it out.

Like it is a sort of magic experience when you go and you've got your design structured in Figma with auto layout. you've got your variables defined and then you're able to just, you know, basically press a button and with MCP, you pull in all that design context into your codebase. Sure.

And you can use inference to basically go figure out uh how to map it. And I'm not saying it's all one shot, but like with either oneshot or a little bit of it or a prompting, if you have your design set up in a structured way, yeah, it it's a pretty wild experience.

you know, Coinbase was telling us about how it's really improved their workflow and um you know, we've heard from many other customers that I'm not sure I have permission to name, but uh how much it's affected them and how powerful it's been. And overall, yeah, people are super excited.

We've been thrilled with the response. That's why we're investing so much time into it is because we really think this is a super important part of our story as a company is being more open and making sure that we really extend out to the greater ecosystem of AI.

Um there's so much that can be done in so many places and design is context and that context is valuable in so many different spots.

So I should think of of the MCP efforts as a almost like a business to consumer like a consumer benefit not necessarily just something like an API that that enables like better uh like B2B partnerships and integrations. Is that roughly correct?

Yeah, I think it's um uh you know I think everyone's still trying to figure it out for sure. Uh but I think that yes um with the right MCP services out there the ways that they can affect um you know all sorts of places that you consume inference uh it's pretty pretty big and I think we're just kind of at the start.

So yeah for Figma uh we're very very excited about where we can go here. Um and then yeah I think just you if you zoom out the bigger story that is being told is uh how do we go from idea to product and you know it used to be this very linear process but now we're no longer in that world.

uh we're in a world where instead uh people try various directions, various paths and you need to be able to generate a bunch of different ideas uh explore them with your team, figure out that option space.

We hope Figma make can be a big part of it, but even if someone has a different tool that they've they're using, um we also want to make sure that they're able to bring their design context in. And um with that design context really help uh that other surface improve.

We think that's a win-win for everyone and it's most importantly a win for the user. I'm so excited for today's middle schoolers and high schoolers to have access to these tools. Like, can you imagine if you had had Figma make? It's incredible.

Like, when you at at that age, like you you you let somebody marinate with these kind of tools for a decade what they're going to be.

I mean, I'm pretty sure we've already seen posts about like a 15year-old sold a company for like $50 million or something like there's crazy stuff happening all over the place, but it's just going to be more and more and more. Thank you so much for joining. Last question. It'll be wild. Last question from the chat.

Uh, do you miss the ratty? Oh, uh, not really. No. What is it at the university dining hall? Is that right? One of them. Yeah. I was more of a V-dubb guy, but uh Okay, there we go. Yeah, the uh uh the was um a great place to see people. Okay.

And but then again, I also like I I hate standing in line, so like I would time out everything so that I either arrive super early or super late so that way I wasn't in the long line. Um but yeah, the food uh maybe not the best, but you know, good the vibes, right? The memories good vibes.

Well, thank you so much for bringing good vibes to this appearance. Yeah, massive week for you guys. Massive week. We'll talk to you soon, Dylan. We'll talk soon. Have a good one. Good to see you. Uh quickly let me tell