Nik (00:05)
All right. Welcome to AI and Design where we explore how artificial intelligence is shaping the world of design. I'm Nick Martelaro.

Dan Saffer (00:12)
And I'm Dan Saffer, and we're faculty here in the frozen winter tundra of Carnegie Mellon's Human-Computer Interaction Institute here in Pittsburgh, Pennsylvania. Each week, we break down the latest AI developments, dive deep into topics that matter to designers, and talk with fascinating guests who are right at the intersection of these fields.

Nik (00:36)
Whether you're a designer working on AI or an AI practitioner interested in design, we're glad you're here. On today's episode, we'll be discussing Google's AI Studio getting a design mode.

Dan Saffer (00:48)
From Davos and the World Economic Forum, some AI bigwigs discussing the future of AI.

Nik (00:57)
and Anthropix's Constitution for Claude, a foundational document that governs how the company's AI assistant thinks and acts. It's a philosophy-heavy guide that even entertains the possibility its AI might be conscious.

Dan Saffer (01:08)
And we'll wrap up by talking about software that is too cheap to meter. When software becomes super cheap to create, people could just have their own tools and apps for things. And what does that mean for designers?

Nik (01:23)
But first, let's get into this week's top AI and design stories, starting with Google AI Studio. Dan, what's going on over at Google?

Dan Saffer (01:31)
there is a rumor that the Google AI Studio

is preparing to introduce a design mode within its AI studio. So that will let people directly modify user interface elements. And this is going to be like a new button on the prompt bar, which activates a cursor tool. And then you can just select any component on the screen. And once that component selected, you can open up an edit toolbox and adjust.

color, font, borders, other properties, just like you do in Figma. what do you think about this, Nick?

Nik (02:06)
So I think this is really interesting to see because a lot of different companies and players have been starting to basically figure out that people want to be able to change things by hand or they want to point and click. They want to say, ⁓ this thing is not in the right place. I just want to move it. Like, I just want to move it a little bit. I want to center this thing. ⁓ just change this text. Just change this color. So the fact that Google's doing this in a tool that primarily was a very much a vibe coding tool, right?

you prompted you speak with Gemini and it builds your whole app for you. mean, actually in in the build studio, usually the code is just in a separate view. You're not even seeing the code generate it. You can if you want, ⁓ but typically you are working with chat and then interface. There's a bunch of people who are doing this now. Cursor has this. They actually implemented a tool in their new browser mode at the end of December.

Dan Saffer (02:47)
Right.

Nik (03:00)
⁓ And I actually know some folks, a former undergraduate research student of mine, Quentin, has a new company that's doing something similar. It's called Inspector. And it's the same idea. Figma as well. Figma has this, they've actually had a feature like this for four months where you've been able to go from Figma Make, their sort of vibe coding tool, and then bring it into Figma so that you can work with actual proper layers.

Dan Saffer (03:23)
So it's funny, we're kind of seeing people coming at this from two different directions. You have Figma, a design tool, moving into the vibe coding space. You have vibe coding tools now moving back into the design space. So think people are realizing...

that there is a value to being able to directly manipulate these tools that it's so much easier to just grab a button and move it three pixels over than do have to explain. Please move that button three pixels over. no, I actually meant four pixels. the classic HCI debate of direct versus indirect manipulation we're seeing now.

playing out writ large in these tools and I personally love it that a lot of these tools are headed back to more direct manipulation so they don't have to know the exact magic password in order to make this thing work

I love it.

Nik (04:23)
I agree. And I think this is something that people of course are going to probably want to use a lot because like you're saying, it's really hard to sometimes put in a language. What is it you want to change? How you want to make something actually updated? Sometimes you can. The interesting thing about it, at least some of the ways that these tools work though, is that you make the change in this browser window. And then it writes a prompt.

and sends it to the code agent. So it actually does the thing where you move it and it will write a prompt that says, move this element. And that's actually one of the ways in which it's doing its update because it's not always a direct manipulation of say, I'm moving the interface and it's updating the code. It's actually move the interface, work through the code agent to then have the code agent update the code.

Dan Saffer (05:08)
That just sounds insane to me, but I get it. I understand why they're doing that. That's how they built the thing. That's what the tech stack is all set up to do. And so they're like, well, we'll just route it through there.

Nik (05:22)
I actually think it's super clever in a way, right? And it's actually a really cool thing to do. The other thing that's going to be interesting about this is that this potentially is a way to get code agents to start to learn So as you start to work with this, as you do manual manipulation, it'll actually then write prompts to then

update the code, but if you might say, maybe it doesn't do it right. And so you might reject that you can actually say, no, that didn't work. You know, when it actually finally updated, I might have to do it again. So in theory, it can kind of start to, learn these systems might be able to capture that data and then start to self-improve. The other thing too, is that as we do more direct manipulation, depending on your service provider and the plan you have, that's going to be more designers giving more, creating more things, updating them, adjusting them.

that's more data for these systems to potentially use to then, to make their initial code generation systems better. Basically over time, they can start to get better and better. And so you may have to do less of that direct editing, which I think is really interesting. Now, of course, it's going to depend on how much data the companies are actually capturing from that. But I think it's a potentially interesting data source if that is what they're doing. I'm not sure if that is.

Dan Saffer (06:19)
Mmm.

so we are basically training these tools to replace us once again.

Nik (06:46)
Well, again, I'm not sure if it's, replacement or if it's simply changing, the position and the role that we're in, right? Are we doing all that direct manipulation ourselves? Are we letting these systems create the initial drafts for us? Do we then go in and edit? And then are we simply then curating stuff? I'm not sure. I mean, I think there's always a risk there.

And I think we need to think about how it is that we incorporate these tools. Of course, if you're a professional, I think you need to probably look at your service agreements, actually look at what they're capturing, what they're not capturing and think about the implications. But I think kind of overall, right? Yeah. It does bring in a question a little bit of where is it on the, creation pipeline. Are we operating on that directing role, which is kind of where a lot of us are moving when we use these tools.

Or are we still working in this kind of middle space where we're having it generate a lot and then we're going in and fine tuning a lot.

Dan Saffer (07:34)
you and I have talked in the past about the shifting role of design and where that's going to land us. And are we ending up with these new kind of roles as product builders where we are doing a lot more of what was traditionally the

the developer's role and where do we sit in that now? we designing prototypes? Are we doing the full on production level things? I hear different things from different places where it's like, I never want designers to be touching production code, one place says, and another place...

which is a big name that don't think that I can say is like, yeah, we're having designers ship production level code in an area where we don't have a lot of developers. And the designers are now just fixing code and shipping small things, not big things, but small things, straight to production. And it's kind of an amazing shift.

Nik (08:33)
Yeah, I've heard stories of designers basically being asked to fill JIRA tickets to update certain small features. And actually, in many ways, I think the tools that we're talking about here, these direct design mode edit tools that would allow a designer to go in, work in a way that they're very comfortable with, which is, drag, drop, add a comment, say, hey, fix this thing, fix this thing. And then it regenerates the code. You inspect it. That's the beautiful thing here, right? Is that

because it's functional, you are experiencing what the end is gonna be, which is different from how we work in the past where we might go in and we might get a Jira ticket to say, okay, we need to update this feature. I might go in to say a Figma file, I'll update that. I'll then send it to my developers. The developers will then do that. So the fact that a designer can do this and now basically with tools like these direct edit mode things, I think it could basically enable a lot of designers to start.

working on production level stuff to whatever degree their company is comfortable with and to whatever degree, their systems are in place to check and make sure it's not going to break stuff. But yeah, I definitely think that we are going to see more of that.

Dan Saffer (09:41)
All right, that is our first story and let's talk about our second story. So while most of America was preparing for a giant snowstorm, the bigwigs around the world were headed to Davos for the World Economic Forum.

including a lot of the heads of some of these AI companies including Demis Hassabis the head of Google DeepMind and Dario Amadei the CEO of Anthropic and the two of them sat down for what was supposed to be a talk about what comes after AGI but they didn't actually

really get to what comes after AGI. was more like how do we get to AGI? What's going to happen when we get near AGI? How will we know if we're going to get near AGI? I don't know about you Nick, but I find that a lot of AGI talk is... Oh, I don't know. It's a lot. It's all prediction. It's all conjecture.

Nik (10:51)
Yeah, I mean, it's interesting to see that, the heads of these companies and this is this topic is brought up at the World Economic Forum. It kind of gives you a sense of how big AI is within the zeitgeist and also how much impact it potentially could have. I think for me watching it, I was thinking about what does this mean? What were some of the things that people were saying that might be interesting from the perspective of a designer?

And one of the things that I thought was super neat was the fact that they were talking about these self-improving loops. specifically, ⁓ Dario Amadei from Anthropic was talking about how what they are trying to do is build a system that writes code and does AI research to then improve AI. So it's this self-improving loop. And I was thinking about this from the perspective of, okay, what does this mean potentially for design and.

Dan Saffer (11:33)
Mm-hmm.

Nik (11:42)
the AI that we use for design and how do you think about sort of self-improving systems that do design? What is required for that? In the space of AI, there are a lot of things like benchmarks. There are these standard metrics that they're going after that they can improve on and they can actually show that they're winning or they're beating another company. Within design, it's a little different, right? Because what is good design?

is a super contextual question. There are some attempts at this. Actually, design arena is kind of like a benchmark for front end design. I don't know if you've seen this, but basically they have the models try to create stuff and then people rate which one is creating better design. It sort of crowdsourced. ⁓ I think this is going to be something that.

Dan Saffer (12:24)
Mm-hmm.

Nik (12:28)
I'm not sure it's going to work at like a global level because design is constantly shifting. What is good is there are some, there are some principles that are always sticking around, but things are always changing. The thing that I was thinking about though, was for designers and for your organizations is potentially how do you build a self-improving system within your own AI design operations?

How do you basically get your AI systems working better, getting closer to what you want to produce, getting closer to what your company wants to produce, and how do you build systems that do that? That was the thing that I took away from this talk, that I was like, that's really interesting, and potentially thinking in that way might be impactful for designers.

Dan Saffer (13:06)
the other thing that was interesting was Debby's, if I can call, if I can use his first name, was talking about the same thing you were, like subjectivity, where sure, we can have these metrics and we know that, yeah, it can solve these math problems

and do these kinds of science problems and can take the bar exam and stuff like that. being able to do subjective work and design is subjective work is something that is very hard to measure. It's very hard to figure out when it's doing it well. And the real problem that they pointed out, which I agree with, is it's not just about

doing the production execution work, it's about asking the right questions. It's about how do I know that this is the right thing to even be working on? And now they're of course saying, well, hey, I'd be able to do that one day. But I think that that's a huge space for design to still be in. you can throw all the AI.

at it you want and it's not going to help if you're working on the wrong thing. So I really liked that kind of framing of well, reframing. That's all about like, how do we know if we're really doing the right thing? And I think that that is something that AI is very far away from. And I think that that is 100 % what they were saying too, which definitely heartened me.

Nik (14:31)
The other thing too, is that you can, if you have these tools that may be on the production side on implementation, if they allow you to work faster, this lets you then find more problems to work on and more of the right problems to work on. mean, there's a lot of stuff where, we, are limited by the fact that like, it just takes time to solve this problem and to do it right and to do it well.

but I do think that these things can improve those iterative loops. So even if we just get that, like I'm actually super excited by the fact that even if we just get a faster iterative loop on solving things where we can implement well, that then we can spend more time on figuring out the right thing to work on. So hopefully we can find and solve more problems.

Dan Saffer (15:13)
Agreed.

Well, speaking of Anthropic, our last news item is that Anthropic released its constitution for Claude, it's AI, talking about

how it should operate. a couple months ago, this document or version of this document leaked and people started calling it the sole document and this was right before Christmas. But now they publish it. It's about 30,000 words or something like that. It's a really long, complex document that is basically a

I don't want to say a rules document, but it is a combination of ethics personality. Operations safety policy manual

I thought it was interesting, and I thought it was interesting particularly from a product point of view. Because I think it starts to differentiate Claude from Gemini and...

CHET GPT and other kinds of AI in that it is starting to have a defined set of characteristics that it really works with. And that's why it has a different tone. And I think a lot of people have noticed that Claude has a very different tone than the major AIs out there. And I think this constitution is one of the reasons why. And I just found that

from just a product standpoint, really interesting because that's a moat right there. now they've released it. So anyone in theory could take that same, document make their own Claude. But I think that having something like this that really starts to define what the personality is.

I think is going to make all the difference. do you think about that?

Nik (17:07)
Yeah, I mean, so we've been talking about definitely in the research world, but even in professional practice, the idea of personality design for AI agents. And I think that this constitution from Claude really kind of steps this up. Like you said, this document is huge and it goes through so much more than I think maybe what we've been

thinking about in the past when it comes to personality, because it is diving into elements of safety, elements of ethics. ⁓ You know, they're putting in a set of guidelines, but they've also written this in such a way where it's like, well, but these are guidelines. They're not hard and fast rules that you have to follow. they basically talk about it as how can we write something to make Claude be a good person?

Dan Saffer (17:35)
Mm-hmm.

Nik (17:51)
And what would a good person be? And in a way I look at this as this is potentially the Anthropics team's way of designing kind of how quad should be as a, if it was a person. and I don't know, I think that that's, it's interesting. I think that maybe this is also the evolution of say personality design in regards to AI agents or chat bots and things like that where.

Dan Saffer (17:51)
Right.

Nik (18:15)
you're moving away from maybe just like, okay, what's the tone you're gonna have? What's the sort of language that you use all the way down into like, no, you actually have to try to almost teach it and give it this way of thinking. And I use that loosely because there's a question if these things are thinking or not, but you you're designing in a way that's sort of like, how does it really try to think, especially in all kinds of different situations?

Dan Saffer (18:39)
Yeah, it reminds me of my college era class that I took in philosophy that was all about how do you live a good life? And there were lots of competing philosophers about this. I'm pretty sure that it was Aristotle who was like, you can't create a set of rules.

because there will always be things that would be evil to do and they're still in the rules it's like sometimes it's better to lie. So you can't have a rule that's like never lie. And so in that sense, I think it's really fascinating. And my question here was,

How do designers get involved with designing these and being in the room when these discussions are made? Because I would think designers and user researchers, people that are talking to customers and thinking about the different use cases and products and stuff would be extremely valuable to help shape these kinds of

personality slash ethics documents.

Nik (19:42)
I, do wonder that as well. I wonder how, designers are going to get involved? But also who else are we now bringing into, into the design process for creating?

these types of documents. think this is really interesting to then bring in philosophers, bring in ethicists. I'm pretty sure that Anthropic does. I believe they have people on staff who are trained in these types of things. But this is the kind of thing, yeah.

Dan Saffer (20:04)
The woman

who spearheaded this, the woman who spearheaded this has a PhD in ethics.

Nik (20:12)
Yeah, exactly. Right. So

this type of document is the kind of thing that really needs a lot of team interaction. And this needs to be something that I think has to be human designed. I imagine that they are trying, they did mention in the document that they wanted.

to eventually be able to come up with similar principles. So they're thinking about this in this way of like weird training and developing the system such that it could also come up with this. But the reality of all these kinds of ways of thinking is that you probably have to have a lot of people in the room. And so this is a really interesting space for thinking about how do you craft a design team around this? How do you craft the team?

Dan Saffer (20:49)
it's funny thinking back to the Asimov three rules of robots and what was that like 50 words maybe compared to the 30,000 that we have for Claude

Nik (21:02)
Yeah. The last thing I'll maybe mention about this is I do say if you do take a look, read the constitution and check out the blog. I also appreciate sort of the humility that the team seems to have. know, they say like, we're probably going to get things wrong. We're probably going to need to update this. I think that that actually is a very sensible way to probably approach this. mean, no one knows really how to create these yet. And so it's cool to see that. I'm going to be interested to see the journey.

and see the changes that the Constitution goes through, basically how it gets redesigned over time.

Dan Saffer (21:33)
It's an insanely hard thing to do. Especially when you have a product that is so malleable. And with bad actors constantly trying to break it and find things for it to do that it shouldn't do and all those kinds of things

Alright, and our last topic of the day is software too cheap to meter.

Nik (22:05)
Yeah. So this was an article that I had read by Steve Newman. And the whole idea was basically personalized software could be something that replaces the kind of big software that we have today. The main idea that I got from this was if you have AI systems that can write code, they can create interfaces. They can create.

Dan Saffer (22:05)
What do we mean by that?

Nik (22:29)
basically full stack software for you and you just have to ask for it in natural language. Does this mean that we can basically create way more software? We kind of know that, but that actually people can create much more personalized software, software that's going to work just for them. And actually in the article, the author talks about creating a piece of software that manages their spam and it just managed it in a way that they thought was better.

than sort of the way that Gmail did it. And so, the software actually, it's not particularly pretty and it might not even work for anyone else because it really is fit for the way that this author was thinking about things. But it works for them. And I thought that this was super, super interesting for two reasons. The first one was what does this mean for design? We spend so much work.

as designers, trying to understand our users, especially a broad swath of our users to create software that works for everyone, or at least works for all of our users. Whereas now you could basically have software where people, they just make stuff on their own or they could have modules and it could just automatically create things. This kind of gets into our generative UI discussion a few weeks ago. But the other thing that it made me think about was also what does this mean for the software that we use as designers?

Right now, we also are users of other software from typically large companies who are doing their best to create awesome stuff for us to build with. But actually like we could start creating our own software and our own tools to do things that work for our own process. So those were like two things that I got thinking about this. Dan, I'm curious when you read it, what did you think?

Dan Saffer (24:06)
at your urging, I went back and this snowy weekend decided I was gonna do some, vibe coding and build a little thing for myself. And so I did, I built a thing that turns all my AI newsletters into a daily podcast.

that I can download, I can generate, download and listen to, makes an RSS feed, I can listen to it and it works, it works fine. Let's put it that way. It did take me roughly 14 hours probably to put it together. And a lot of that was resting the terminology.

wrestling with a whole bunch of new systems that I'd never used before and some of it was Yeah, like I had to buy a whole bunch of new services and I had to do all so as I was doing it I was like there's no way that People are gonna do this as is

I think that there is a small group of people that yeah, 100 % they will use these tools and make them. And if you're going to do something super simple, sure. Like if you want to use it to make your website, stuff like that's probably really easy and really good. If you want to do anything beyond that, like if you're talking about, I'm going to make a custom design tool.

Sure, I think that there is going to be a group of people that does that. And there's a group of people that are doing that right now. There's people that are building agents to work on different parts of their design process. And that's super cool. It's also something that I doubt that we're going to see being pushed out to the world. There's all kinds of things that I could do myself right now.

But I'm not gonna make my own furniture. I'm gonna go somewhere and buy it. I'm not gonna build my own car, even though I probably could, or all those kinds of things. You could make your own electronics. And I just don't see people building a lot of custom software.

where they're thinking about it as a software thing. I could see in the not too distant future someone saying, hey, make me a podcast of all my, newsletters today and feed it to my.

things and have that work where I'm not having to think about putting it all together and releasing it and distributing it. I could definitely see the more generative like, it just makes me this little app that I don't want to think about it. But we are not close to that now.

Nik (26:57)
So I'm excited to hear that you tried it and you built your own thing. That's super cool. Yeah.

Dan Saffer (27:02)
It was fun. It was fun. Don't get me wrong. It's

a little like a code slot machine where it's like, okay, let's try this now. And I had forgotten with my very limited code background, how much of coding is actually just debugging. And it is just an incredible amount of debugging. Oh, that doesn't work. Oh, this doesn't work. Oh, you know, just to get something. Like last night at almost midnight, I was like, yes, I finally got it. And it was...

It was great. But yeah, there were some frustrating hours that I admit that had we not been doing this podcast, I probably would have been like, screw it, I'm not doing this and walked away. But I'm glad that I did. I'm glad that I pushed through.

Nik (27:46)
Fair enough.

So there's a couple of thoughts. mean, I want to say, I think you're probably right in the near term that the average person is probably not just going to hop into one of these vibe coding tools and say, make me this thing because you're right. I've done this. I've been, I build stuff all the time with these tools now. and even with, I have a bit more of an engineering background. I'm still spending hours and hours now for me, actually it's.

It's fun. Honestly, it's been great. I love spending hours kind of, you know, locked in doing this stuff. That being said, that I think the average person is not. The average person though might have something in their life that they're willing to potentially do this for. And the question I have is what interfaces, what interactions do we need to design for them to say, take it from one level or you've got you and me.

We're in the tech field, we know this stuff a little bit or a decent amount. What's like the next level? I'm trying to think almost like what's like the hypercard of today? Because hypercard did a lot for making it so that people could design effectively their own software. mean, there's people who still miss hypercard. And I feel like in many ways, we still don't even have some of the capabilities. in some ways we have more. But yeah, I kind of wonder like...

Dan Saffer (28:59)
Sure.

Nik (29:07)
What is it that we need to do to design something that just, it moves it up for someone with an ability that's a little less than, someone who knows how to build software, so they don't really know how to build software that they can do this.

Dan Saffer (29:18)
What you're really talking about here, it sounds like, is like personal product market fit where an AI recognizes, hey, you have a problem in your life. I can fix that. in my case it's, hey, I've noticed that you have a lot of AI newsletters that go unread in your inbox.

How about if I make you a thing that just delivers a podcast to your ears every day for 10 minutes? Cause I've noticed that you listen to a lot of podcasts. That would be awesome. But that is,

a very smart AI that knows a lot of context about you and know where you struggle.

and they know where they can apply some technology to help solve that problem. They're like a personal product manager or a personal user researcher or a personal designer all rolled into one. And I think this goes back to what Demis was saying about AI not being smart enough to be able to guess those problems yet.

If you said hey, do this for me, It could help you do that. I mean, that's what they're all doing now. That's what cursor does or or replator called code or you know that's what all of them do now.

But you have to approach it first. You have to be the one who's like, hey, make this for me. And A, that requires having an understanding about that it can do that, which I don't think a lot of people do have that outside of a small number of people. Two, you have to then be able to like massage it for 14 hours to get it into the state that you want it to be. And I think those are two pretty big hurdles to get over.

Nik (30:55)
Yeah. I mean, I think that what we have right now are systems, are AI software developers that have some skills in say design and the ability to do design craft. What you're talking about is I think closer to what designers really do, which is understanding a problem space, understanding users.

building prototypes to learn, right? Building things to test and experiment, not just to say, I want this, here it is, here's what this does, but saying, it could be this, it could be this, here's three options. Actually, this is the thing. None of the systems that I've seen have ever really given something where it says, hey, I'm not totally sure what you mean, but I went back and I made three options. Can you just try these out? Tell me about how are you feeling about this one? Is this what you're thinking? Is this what you're thinking? You might not even know, that's the interesting thing.

When we work with people, they don't even know what they want. And oftentimes when we do design work, we might not even know what we want. This is why we go through this iterative production, because we're figuring it out as we work with the material. This is something that I haven't seen a lot of in regards to how these tools work. Maybe we'll start seeing more. I'll say too that we also haven't seen a lot of the tools acting like this design will. Now, full disclosure for folks listening.

Dan Saffer (31:53)
Right?

Nik (32:10)
I run a research group here at CMU that is quite interested in these questions. I myself am interested in these things of like, what if machines could start understanding user needs and how would they do that? What processes might they do? How would they basically start defining problems? And really, how would they work with you to do this? And so that's maybe something here where I think we're getting to a point. The software, the coding is basically becoming too cheap to meter. The design work.

to get you to the step of creating software is still really expensive, either in regards to time or regards to people, our effort. So a question there is then what are the systems that you could potentially design that maybe start doing that?

Dan Saffer (32:53)
will say that one thing that's cool about this is you can take someone's idea that is so small that no one is ever going to commercially release this thing. But there may be a small number of people that would pay for it or

take it and make their own version of it. And so that's really cool because there's no way that Figma, for example, is going to make a design tool that does not work for 20,000 people or 50,000 people. They're just not going to do it. They're going to homogenize it in ways both good and bad.

for a broad audience to get a lot of people working on it rather than hey I made this tool for for myself if you need if you like it great now I do think this has some drawbacks in that one of the things that is nice about having a community of people around

Certain pieces of software like there's a community around figma right now is that you can help each other and people build stuff on top of a platform and People make plugins and make templates people do all these things and it becomes a

Community, becomes a way of sharing knowledge that you're probably not going to get in a very fragmented... If everyone had their own version of Figma, you would have a very different... There'd be a very different world, very different design community. Now some may be like, well that would be a great design community. I'm less certain because I do see that it is great to have...

tools that other people have built on top of a tool or a platform that can be shared. I think those are valuable things.

Nik (34:37)
Yeah, so now you're making me think two thoughts. One is plugins too cheap to meter, right? Basically building a system, if you have a tool that is expandable via plugins, tools like Figma, tools like Adobe, right? How do you make it so that you build these AI tools that let you build plugins super fast, super easy, such that

Dan Saffer (34:42)
Mmm.

Nik (34:56)
some idea and they're willing to put in that 14 hours that they can create that for their team. And then yeah, maybe share that out with the community, maybe get people working on it. I mean, this is where you start bringing in, ethics and values of open source. And that, that can be really cool. And, know, some of those ideas could potentially grow. Some of them could also just be great for like you're saying, like the 10 people, but the fact that it's maybe built on say a larger platform, it's not, it's really just the plugin. It's not the whole tool that you're.

you know, I have no support on this. It's just like, wow, we did this. And also you have the support of all these other people who might want to be interested in your plugin kind of jumping in. I mean, that's maybe that's like the one of the first places, or maybe that's a place where some of these larger companies that have tools that are extensible could start thinking.

Dan Saffer (35:43)
Well, the nice thing would be oh, I have an idea for a plug-in but I've never made a plug-in but I can certainly talk to Claude about making a plug-in or Gemini or whomever and make make a plug-in in a way that I certainly couldn't have a Year ago or two years ago, but now that stuff is super possible

Nik (36:07)
And this then brings up another thought that I had around design teams within companies who are acting as the system and tooling designers for the company. So some companies have this. know Google has, people who are working internal and they build internal tools. their users are Google employees. And I'm sure

A lot of companies have this, or at least once you're a certain size. This is an interesting thing though. You could imagine having, for example, like one or two designers maybe equipped with these software tools and the ability to like go around, do internal user research, figure out the processes and say, we're gonna build tools for our own company. We don't need to release these. We're not going to make these.

There's a lot of tools that end up being like their startups that were actually some internal tool that then they were like, actually, this is super valuable for like all the other developers or designers out in the world. And we're just going to like launch this thing. And sometimes that's the thing that's more potentially profitable. Right.

Dan Saffer (37:05)
I mean, that's what Slack was, right? Slack

was a side tool. And yeah, it turned out that We all needed to search for messages that we couldn't find on Slack all day.

Nik (37:17)
Ha ha!

Dan Saffer (37:19)
Pixar famously has software tools that are just for the animators

Nik (37:24)
Yeah, there's actually this article called an app can be a home cooked meal. This is from an author, Robin Sloan. It's from 2020, I believe. this idea was the article here was this idea that like we could build apps just for ourselves or for like our family. And this was 2020, right? This is pre all vibe coding tools. So the now, I mean, I actually think that again, you can do this. I think.

Dan Saffer (37:32)
Mmm. Mm-hmm.

Nik (37:50)
In this case, the idea is more like, you are a developer, a designer. Like, you know how to build software. You could build software for like your family. And in a similar way, right? Like you can do this now for your company. You can do it for your team.

Another example might be, as I'm thinking more about where could this be really powerful is things like community organizations, nonprofits. Like if you're someone who likes to support certain groups, one way to support of course is through donations and other ways through donating your time. But oftentimes, many of these groups, like they don't have a lot of technology. They're not online. They don't have these things. may, who knows, they need stuff. I don't know. Maybe this is like a way also where like you can utilize these things.

Because one could argue like, is it that are they gonna learn it themselves? Maybe, but they've got other things to do. But like maybe you as someone who knows enough about this, right? It's still that middle person, right? You know enough about tech and now you know enough about with these tools. I don't know, that's kind of cool. I just, this is also making me think about that. Like even moving beyond just like internal tools or tools for the Ham Family, like tools for organizations you care about. Yeah, that could be interesting.

Dan Saffer (38:53)
It still comes back to understanding the problem. It's that you are building the right tool. otherwise I think a lot of these things are just going to sit on the shelf. they're going to be built. that was fun. That's cool. Never use it again. There'll be thousands or millions of these one-off apps that are just sitting on GitHub.

collecting digital dust and that'll be because even if it's just for a single person it may not meet people's needs and I think we're even bad about knowing what our own needs are sometimes so it's so having something that really works within your life and lifestyle and

all that stuff is really hard and designers and researchers and stuff spend a lot of time making that stuff work. That's what makes some of these best apps really good is that people spend a lot of time conceiving them, building them, testing them, refining them, rethinking them, and over time

That's how and why apps get good. especially pushing that all into one person makes it really hard.

Nik (40:13)
Yeah, that's actually super interesting, the idea that it gets pushed onto a person rather than a team. Because it also means that, how much of something is going to be limited by the thinking that that one person has. I think when you're thinking about it in this home-cooked meal type thing or an internal tool, that's maybe okay. The one person helping a few people, that's kind of all you got.

But for example, if we are thinking and maybe pushing towards this vision of like everyone could be developing software all the time and pushing it out and then people are trying to use it, that could be a problem because you are limited, you're limited in your way of thinking. That said, right, I do wonder, and again, this kind of comes back to the idea from Newman's article though, like what if it's just, no, no, you do end up with some type of systems that can...

basically be a designer, there's an AI designer in there somewhere that is designing for you and is like, no, just software, it's for you and it only needs to be about you.

Dan Saffer (41:17)
and all watched over by machines of loving grace. That's the dream,

Nik (41:21)
Yep.

Dan Saffer (41:24)
All right, that's a good place to end. Next week we'll be back talking about user research in the world of AI. Where do user researchers fit in?

Is there a place for them in a world of fake personas and synthetic So we'll be back next week to talk about that. And stay warm, everybody.