Design in the Age of AI: Craft, Tools, and the Future of Creativity

0:00 Vince Dean: I haven't seen like craft, I've seen like sort of an amalgamation of the deductive whatever it is that spits out, but like really figuring out a way to draw craft out of that. I don't know that it can do that. 'cause you know, our jobs are sort of. To bring out the human factor, and whatever you're looking at or whatever you're using,

0:38 Brad: All right. Welcome everybody to another episode of the How Many CTOs Take podcast. I'm Brad Hefta-Gaub and my co-host is. Scott Porad. And today we have a very special guest with us, uh, Vince, Dean, Vince, uh, and I go way back. Um, and, uh, we're gonna talk to Vince about, uh, design and, and what the, what the latest is happening there.

1:06 Vince Dean: Sure. Well, um, Brad and I of course have a lot of history. We. I think we started working together in 99. It's been a while

1:14 Brad: probably.

1:15 Vince Dean: Um, but, uh, I, you know, I kicked off my design career. More than two decades plus ago, and, and been in and outta corporate since.

2:33 Brad: very cool.

2:33 Scott Porad: Quick, can I double click on the connected TV stuff for a second? How, what ways was it similar to the old time internet in terms of being difficult?

2:43 Vince Dean: It really, when we started, um, it really felt like the wild, wild west, you know? Um, it's 10 years ago when I got started, and now it's been very much like systematized and template templatized.

3:47 Brad: can imagine that, uh, kind of when you say the wild west of the early web, you know, I kind of picture back when, uh, it was, you know, mosaic and Internet Explorer or something like that.

4:32 Vince Dean: Yeah, it's gotten much more standardized by today's standards. I mean, the last project I did was for PBS. Um. It's actually for the local conglomerate here in town called Cascade, PBS.

5:15 Scott Porad: I'd like to come back to the navigation point for a second and get your take on something you mentioned that. At the beginning of the web how navigation was and websites was all over the place and it's standardized a little to the top and the left.

5:49 Brad: Aren't we all?

5:50 Scott Porad: Yeah. And in the family, you know, I have some folks that are a little bit older. My, my parents are still around. I've got in-laws, et cetera. And believe it or not, um, I had some grandparents that lived into their nineties. I had four of them actually, actually I had two, and Lisa too. And so I was even doing sort of tech support for folks who are really old and uh, every time I have to help them, especially on mobile apps, it's so obvious to me how inconsistent.

7:11 Vince Dean: Yeah, I mean, I, I agree with you on standardization, but every once in a while a disruptor comes along and changes.

8:27 Brad: yeah. I mean, it's interesting, Vince, you know, one of the things that I have always, uh, valued about working with you, and I think this is, you know, I don't, you know, uh, you're, you're quite good at what you do and there's plenty of people I've worked with in design that are good at what they do and, um.

9:38 Vince Dean: or just like respect across the board. You know, a designer should respect, you know, what the engineers are doing, and the engineers should respect what the designers are doing. And, and it goes all the way to the top. There needs to be some sort, sort of synergy there that where, where you get good results and, um, you know, when you work in an organization where people, you know, don't partly understand and respect each discipline.

10:04 Brad: And you're getting better stuff. Yeah, yeah, totally. Um, so the, a little bit of a hard pivot, but I'm kind of curious, uh, Vince, to get, I mean, maybe, maybe the take on this is gonna be interesting, like this, uh, the world of LLMs and this sort of like, churning of everything that's ever been built, you know, every design that's ever been done, you know, existing in this like, space of, uh, reproducibility of, of generative ai.

10:41 Vince Dean: It's changed my sort of approach quite a bit. Um, but, uh, and, but the principles around design are still the same. Uh, I mean, I've been really heavily experimenting with AI for the last couple of years and, um.

13:39 Brad: right? Do you see that? Same? So what I hear you saying is kind of, it sounds very similar to what I have experienced in the, in the side of sort of writing code where it's like, um, you know, I come to the table with a.

14:46 Vince Dean: Yeah. I wanted to, like, for a designer like me and many, many designers that I know they have. Thousands of Figma files and thousands of things within a Figma file and layouts and icons and all of those things. And I simply like, I sort of bought into some of the hype around NCP. Right? Um, and I forget what the acronym is for

15:09 Brad: Think it's model context Protocol.

15:12 Vince Dean: Um, so. Essentially it, it's a, it's a tool that will allow you to connect very, your applications, your local applications to an ll and m. And it goes out and, and does that through a repo like GitHub and you can have it check things, right? That's the way I understand it. I'm a complete novice, by the way.

16:21 Brad: Um, and, and to clarify, Vince, 'cause uh, so this is, you had this idea of a type of common action that you have to do as a designer.

16:35 Vince Dean: that's a big thing, right? Like, is the copy legible on this gray background? Let's have an LLM check that. There's no easy way to do that. Um, so what I wanted, ultimately what I wanted was to have an LLM look at all my Figma files and uncover where ex accessibility was an issue.

17:14 Brad: it's interesting right? Because you, I, what I hear in that story is the idea that as a designer you have very real workflow things that you have to do.

18:07 Vince Dean: Yeah. It, it failed. Like, and, and, and I will say maybe I did something wrong, but I did get them talking to one another. The, the, the cool thing that came out of that project was, is. And I, I use Gemini. Gemini quite a bit. I use all of them, but Gemini is really good at being a partner at troubleshooting something.

19:26 Scott Porad: So, can I go back to this accessibility question for a moment? Sit sitting aside the AI part if we wanted to.

19:48 Vince Dean: yeah, there is, there's actually a tool in Figma that will do that, but it only checks incrementally that, that the things you tell it to check.

20:20 Scott Porad: Yeah, so like Figma doesn't have like a check all button

20:24 Vince Dean: if there is one.

20:26 Scott Porad: Right. I, I don't know. I actually have a, can I ask, can I detour for a second? Sidebar for a second and ask about Figma?

20:31 Brad: Mm-hmm.

20:32 Scott Porad: So look, I'm a, I'm a, I've worked in engineering and product management for my career. I've never really worked in design. In fact, I have a, when I'm introducing myself to a new team, I have this slide I give, which is like, Hey, here's all the things I've done in my career.

20:53 Brad: why, that's why you've never landed a job in design is, is 'cause you

20:57 Scott Porad: always use comic sands. That's it. Pardon me. So it, see, I have a question about Figma and it, this is gonna require maybe you giving us or me a little bit of education about the design tool landscape.

21:39 Vince Dean: Yeah. I The difference between like Figma, the, the, the core difference between Figma and, and Photoshop as Figma was made specifically for product. And so they built all the tools inside of it around building for product. Photoshop is for pictures.

23:13 Scott Porad: Interesting. So they were, whereas an industry we were, we were taking tools really designed for other purposes and trying to make them design online products. Figma, whoever owns Figma came along and said, let's make a tool really designed. For people who are designing online products, designers yourself.

23:43 Vince Dean: Yeah, there were a few. I, like I said, I think it's XD is the other one.

24:01 Brad: Um, in some sense it's like you said, Vince, it's like they were disruptors in that they, yeah.

24:48 Scott Porad: Okay, so now I have a follow up question. As an engineering leader, what can I do? What can I have my engineers do to work with you and the design team and figment, et cetera, to get them to make, make it easiest and most effective for you?

25:03 Vince Dean: Most effective for me is, Hmm, that's a good question.

25:07 Scott Porad: Lemme ask a different, a tangently related question. It does Figma actually produce the style sheet, the CSS code.

25:16 Vince Dean: Yes. Yes.

25:17 Scott Porad: Oh, interesting. Yeah. So, so in the old world where a designer would give an engineer an image and the engineer would try and figure out the CSS from that, in the Figma world, it generates that.

25:28 Vince Dean: Yeah. And it's not perfect by any stretch, but the styles are there.

25:31 Scott Porad: Interesting.

25:32 Vince Dean: Um, and if you, if you design it and build it out perfect, you'll get really good results. But if you're sloppy about it, you won't. Um mm-hmm. And the other beauty of Figma is like you, you can set up a whole design system, like you have these things that you reuse throughout all of your layouts.

26:33 Scott Porad: so you're giving me a clue of, of something we can do, which is to the extent that as engineers, we keep.

26:58 Vince Dean: that's really, really hard.

26:59 Brad: Yeah. That's one of the things I think that, uh, when we chatted with Greg Conklin, that was one of his things. He said early on, you know, like early on. Get a design system, make sure you're using a design system.

27:09 Vince Dean: And the, the other, the hard part about design systems is if you are, say, a startup, for instance, you don't build a design system until you're, you know.

28:03 Brad: you know what that sounds like, Scott? No. What's that sound like?

28:41 Vince Dean: Yeah. I mean, you could be very conscious of that build process along the way.

28:55 Scott Porad: a, if I were starting a, an A project today, whether mobile app, online ct, or whatever. Are there like open source design systems I could just sort of download to start with?

29:07 Vince Dean: Yeah. There are tons of them. Yeah. Oh,

29:10 Brad: mm-hmm. Like skeleton or tailwind. Yeah. Or right. Yeah.

29:15 Vince Dean: And you can take those and sort of modify them to suit your needs.

29:19 Scott Porad: Yeah. But as you're saying, when you're getting started with your application, whatever it is, you might, you need to. Iterate and modify that design, then you're kind of, there's an interesting trade off there.

29:33 Vince Dean: Yep. Yep. And you get to a point where if you have a really good design system, an engineer can just take the modules and create, create a layout on their own. They don't need design and even involved, um, yeah, maybe look at it, qa, you know, that type of thing when you're ready to push it.

29:53 Brad: Um, like the product owner can come say, oh, we now want a mute button here. And it's like, oh, well I'm just gonna go use the button from the button module. And um, I already had a, a, there was already a speaker icon that existed on the volume control. I'm gonna boom. Just drop that in. Now we got a mute button.

30:12 Scott Porad: Well, here's, here's an interesting thing I'm, I'm putting together, which is if from an, from the engineering side, if you're gonna try to get the AI to write your software for you, having a design system is gonna make it better because you can say, here's the design system you use, here's the modules I wanna use, whatever they're called in a design system, and.

30:45 Vince Dean: That's exactly right. And it's getting to the point where you can actually have a layout. And within that layout there are design system, uh, ingredients.

31:42 Brad: I sort of feel like this is another example, Scott, that you and I have talked about where like I've had the good fortune of working with folks like Vince who are willing to be patient enough with me to explain why a design system matters, you know, and help me understand that as a develop, you know, leading a development team, that if I can work within the cons, constraints of this design system, or we can collaborate on a design system that's gonna all of a sudden make us more efficient in the long run.

33:07 Vince Dean: Yeah. The meta part about it is like, you can tell it to, you can tell, tell it to help you know what I'm doing, so you, you know what I mean? So it's pretty insane. Like, and, and the other thing that sort of plays into that, that I've seen like the trend is like a lot of people are building these out with actually problem solving first and like.

33:57 Scott Porad: Well, I, I heard I read something online, tweet length kind of thing, uh, which, you know, a designer question. You know, will AI really, you know, fill designers, so to speak? Not like murder, but like career. The person said, no, you still have to get the clients to tell you what they want. And you know, if they can't tell a human, they're not gonna be able to tell the ai.

34:25 Brad: Oh, that's an interesting, hold on a second. I think I, I think I hear what you're saying, Scott. Let me just make sure I'm getting this. This is similar to what we talk about with engineering, that it's that being that Rosetta Stone between like what the product owner says. They think they, they think they know what they want and they're sort of using words to describe it.

35:03 Vince Dean: Right. I mean, we have to help the, the, the client or the product owner, whoever it is, to distill their ideas into something that's tangible and oftentimes it's wrong and we have to figure out a way to tell 'em it's wrong. So, right. That's the other interesting about thing about AI is you can tell it that, you know, tell it to question my, my assumptions.

35:35 Brad: yeah. That is interesting. I have noticed that the LLMs tend to be, they default to quite agreeable, right? If you tell them you want something, it, they're very agreeable outta the gate.

36:01 Vince Dean: And then locking on, you know, cliches and tropes and all of that stuff.

36:17 Scott Porad: here's a really interesting question about AI and design, which is. You know, the design is a fundamentally a creative enterprise.

37:09 Vince Dean: Yeah. And that's the thing that, the results that I've seen, it's not like I haven't seen like craft, I've seen like.

37:47 So,

37:49 Brad: uh, so what is your, okay. So very much related to this question of like, it, it's just kind of regurgitating stuff, um, and that that true craft, true innovation kind of has to have, uh, can't just be regurgitation of something else. Um, how do you feel about sort of the. Like everybody draw, draw myself as a Peanuts character, draw myself as a studio Ghibli character.

38:40 Vince Dean: Yeah. The implications of that is like, it's, it, it's, it's sort of mind boggling and I, I don't know how to, how to, you know, even talk about it. Um, I mean, I, I'm, I love generating images with Midjourney and seeing what I get. Um, it's really interesting some of the things that spits out, I mean, some of the posts that I've done Brad, um, are all created with Midjourney.

39:25 Brad: I, I feel like maybe this is, uh, a little bit of a provocative statement on my part, but when people complain to me that the LLMs are essentially copying something that it saw, I, I make the point that like.

39:47 Vince Dean: Yeah, we're standing on the shoulder of giants, you know, like that same cult's, true to music and art and whatever you, you know, any art form that you're in. So

39:56 Scott Porad: there, there was a pretty big court ruling in the United States about that.

40:53 Vince Dean: So it, there's so much gray area there. Like what is, it didn't. New York Times just signed a licensing deal with one of the LLMs.

41:01 Scott Porad: Right.

41:02 Vince Dean: And I'm like, really? They did that? And, and, and in my mind I thought somebody like the New York Times just would not do that. So, yeah.

41:11 Scott Porad: Well that's interesting too, 'cause that might have just been purely a business decision from the perspective of, well, they're gonna get our content one way or the other, so we might as well get, we might as well get some money for it.

41:22 Brad: Yeah,

41:23 Vince Dean: why not?

41:24 Brad: I do think it's interesting, uh, Vince, and at the risk of like becoming a little too philosophical, um, I, I appreciate that you as a visual, creative artist, right? And you, and your point is like, this is true for music as well, right? Or or storytelling. Like Yeah. We, as artists, even if we're creating original content, we can't help but be inspired by everything that we consumed.

42:22 Vince Dean: No, I mean, we tap into and get inspired by things, you know, like it, it's every single project that I've ever done that has some meat on it. The first, the first step in the process has always been to go out and look at the competition and just look at examples.

42:42 Brad: The mood board, basically.

42:44 Vince Dean: Yeah. Um, yeah. And I, I love that part of the exercise because you, it really gets the creative juices flowing. You get excited about where you're looking at and it's, you discover new things and, you know, leverage some of those ideas. Don't outright steal them.

43:09 Brad: Right, right, right. And that's those things is what I, I, I can't, I, I remember many times you and I have been in a room either working on a project together or you know, working on somebody else's project together, and it's like.

43:48 Scott Porad: And that's just one more argument that we're on the simulation. Brad.

43:52 Brad: Maybe, maybe. Um, so, uh, Vince, like you do use Midjourney and you're cranking stuff out with it, and you're pushing it in certain ways, and it, and, and yet it, you're sort of feel like it's all vanilla.

44:38 Vince Dean: Well, I mean, the visual stuff, like the stuff I get with Midjourney, it's essentially a picture. It's not product oriented. Um, like the other thing that, that it just cannot do yet is like you get a, you get something that you like, you get a result that you sort of halfway like at a midjourney and then you repro it until it.

45:53 Brad: Yeah. Like you can't ask it to go edit you once you have it. Don't ask it to change it, 'cause it'll just come up with something completely different.

46:01 Vince Dean: And don't get really specific because the minute you start getting specific, it's not gonna give you what you want. It's just not. So, um, some of the imagery's really, really, really cool.

46:21 Brad: but it's very hit or miss basically. Yeah. Uh, so Scott, I'm gonna nerd out for a second. So, uh, uh, codex, uh, chat, GPT Codex, which is one of the agent programming things you can basically, like, it's, it's similar to and GI GitHub.

47:41 Scott Porad: That, that is a funny thing. You can never get the AI to come back with the same thing twice.

47:47 Brad: That's essentially, yeah, that's it.

47:49 Scott Porad: Which is a puzzler, because sometimes you want it to come back with the same thing twice.

48:04 Brad: It, it, it struggles at that for sure.

48:06 Scott Porad: Don't change a single pixel except for these few pixels right here. And it

48:10 Brad: it have you impatient?

48:35 Vince Dean: It's just way more of a crap shoot. And, and the code part, I totally get what you're talking about too. 'cause I, I can't remember the last thing I was working on, but I, maybe it was that MPC thing, MCP thing. I had this incredibly long conversation with, um, Jim and I, and what I found is, and it this has happened before, that more than once is it will undo.

49:20 Scott Porad: we've heard that before and I would love at least in like the coding space, and it might make sense in the design space as well.

49:49 Brad: Yeah, and cur cursor starts starting to have some things, like a rollback to the previous thing.

49:53 Vince Dean: Mm-hmm.

49:54 Brad: Um,

49:54 Vince Dean: yeah, there's like integration with GitHub too, so you could, you could submit it at incremental incrementally as you're fixing things, so it doesn't, I just

50:02 Scott Porad: wanna go to tell the air as I'm talking to it, like, okay, cool.

50:13 Brad: Um, Vince, I'm kind of curious. One of the things that we have when we've talked to other folks that are really into the, um, uh, sort of using, uh, AI generated code from a software development perspective, uh, we often compare it to these like junior engineers and we kind of talk often about, and I know that at least with my team, I'm really coaching them around like, you know, treat this thing as something that you've gotta supervise.

51:17 Vince Dean: using it for explicitly for design, I just haven't really done it yet. Like outside of, and even then, it's not really designed.

53:00 Brad: so it's not right. Yeah, they're not, they're not coachable yet.

53:04 Vince Dean: No.

53:07 Brad: Yeah, it's so fascinating. Like, is it what, so what's your take?

53:26 Vince Dean: Well, it'll be interesting to see and like, I don't know what the, like, there's some weirdness around it because if you, if you have a daily LLM that you use and it actually gets to know you and learns from that and sort of gives you answers based on everything else that it's learned from you, clearly that's begins to be coaching right.

54:14 Brad: Yeah, it's getting there. Well, this was super fun, Vince.

54:45 Vince Dean: I appreciate it. I appreciate it, uh, being on board a lot of fun. Right. Thank you. I love talking about this stuff.

54:54 Intro/Outro: This episode of how many CTOs does it Take Podcast was edited by Saffron Hefta-Gaub. Our theme song was AI generated by Brad. Please follow us on all the social media platforms at at how many CTOs Pod you can find us.