Matthew: Welcome to the Restock Podcast by No Code Supply Co., where we highlight the people, the things they build, useful tools, visual inspiration, and conversations to refill our shelves and ourselves. Welcome and I'm Matthew. Corey Moen: I'm Corey and today we have Liam McCabe, co founder of Code and Wander digital design agency joining us. Thank you so much, Liam, for for coming. You have some rad stuff y'all have been building. So we're really excited to see what you've been up to and yeah, just get to know you. Liam McCabe: Hey guys, thanks for the invite. Looking forward to it. Matthew: one reason we wanted to ask you to come and join us is you know, we on a previous episode had checked out on your agency site the new, well, new to us anyway. I don't we'll we can get into the timeline of this, but there's an ask functionality where you can just kind of query on the site, but instead of doing a site search, it's actually taking that query and then kind of building out some some customized sections. for based on that query. So really just interested in how you built it, what kind of spur spurred this idea on, what was kind of the stack behind it and kind of what how have the results been from that? Liam McCabe: Yeah, yeah. yeah, basically more recently we've been offering a bunch of AI services to our clients. And so Matthew: Mm-hmm. Liam McCabe: we needed to update our website. So with a couple or two or three service pages, just AI services, workflows, automation, all that sort of stuff. Matthew: Mm-hmm. Liam McCabe: And we try to practice what we preach and like, there was nothing really AI driven on the website. So it was like, Matthew: Mm. Liam McCabe: although we have done it for clients, case studies can only get you so far. So we were like, what could we actually do on our company website? And obviously the go-to is like some sort of chat bot, put it in the corner, ask questions about Code and Wonder. But it's like, that's a bit typical, a bit boring and everyone really hates chat bots kind of, even though we're talking with Matthew: Yeah. Liam McCabe: Claude or ChatGPT every day. But Matthew: Mm-hmm. Liam McCabe: having one of those in the corner felt a bit... Yeah, not attractive. So we brainstormed a few solutions. And basically in the past, I've been messing around with UI generation for a while Matthew: Mm-hmm. Liam McCabe: and trying to basically marry a design system with a knowledge base, a data source, and essentially try to generate any sort of page that you want from that. And in the past, maybe I've been trying to extract design systems. Maybe I should have just used an existing design system. Matthew: Mm-hmm. Liam McCabe: But I always struggled with that. So Matthew: Right. Liam McCabe: I always just fell short and was like, I'll try it another time. And then more recently, it is very new to the site. Yeah, I was able to extract Code and Wonder's design system really well, all the tokens and even the markup for the components. And also actually the scripts for the interactive components, which has also struggled in the past to do that. Matthew: Hmm. Yeah. Liam McCabe: So it's done a fantastic job on our website to do that. And so I was like, let's see if we can actually integrate this into our site. Matthew: So these these aren't like hidden components that are on the page and you're kind of injecting the content into them and revealing them. You're actually injecting the HTML, CSS, and JavaScript on demand. Liam McCabe: Yeah, so what you're Matthew: Just in time. Liam McCabe: actually seeing is a React application in an iframe, essentially, because Corey Moen: Yeah. Liam McCabe: it has to be client-side, client-ready, right? Everything is Matthew: Mm-hmm. Liam McCabe: generated on the fly. So yeah, it's basically built a design system in React. It's got all the design system tokens there. And then from there, it basically takes the query, sends it off to Claude. There's a bunch of schema that's actually tied to the design system and what you can use, when to use it. Obviously a page tends to start with a hero. Matthew: Mm-hmm. Liam McCabe: There's a predominant component, which is like a two column layout, which you'll likely wanna use all the time. For others, there might be a grid, list of items, for example, and it's like, well, if you're using that component, then you always wanna use two or four for the grid layout. Yeah, so there's a bunch of rules that comes with that query that's sent to Claude or whatever model you wanna use. And then alongside that is the knowledge base. And that's essentially, for us, it's just an extraction of the website content. Matthew: Okay. Liam McCabe: But it could really just be any sort of knowledge base, really. It could just be a spreadsheet, a database, whatever. And yeah, that is sent alongside in the payload, which works as an MVP in the long Matthew: Mm. Liam McCabe: run. You probably want to use some sort of Ragtabit database or something. Matthew: Yeah. Liam McCabe: But yeah, that's working so far for us. Matthew: Yeah. So you don't have to manage a second database of the kind of the same content or extended content. It's all it's just using Liam McCabe: Yeah. Matthew: what's there on the website currently. Liam McCabe: Yeah, because we've been thinking a lot about the single source of truth data model Matthew: Mm-hmm. Liam McCabe: type thing, but I'm wondering how realistic that actually is. Because Matthew: Mm-hmm. Liam McCabe: we obviously use so many services day to day, whether it's Slack or some sort of notion type thing. And I imagine all of those could sync to a single source, or you could just keep that. sync alive somehow. So yeah, that's Corey Moen: Mm-hmm. Liam McCabe: something we've been thinking about. But yeah, currently it's very MVP, literally a JSON. A massive JSON. Corey Moen: It's awesome. Matthew: Isn't everything JSON these days? Feels like it. Liam McCabe: At the end of the day, yep. Corey Moen: And there and there's I I think you maybe said this, there there is like a model that you're running here that is making the decisions, like take take taking the question from the query and then mapping it to what it thinks the content is that's best. Right. Liam McCabe: Yeah, so I guess you could use any model, AI model to do this. I think I tend to lean towards Anthropic these days, Opus and stuff like that. And then, yeah, we're basically saying, we actually created a little tool, so it's called emit page. And with emit page, it figures out from the query, does this need like a... a single word answer, a paragraph answer, or can we generate Matthew: Mm. Liam McCabe: a page from this? Corey Moen: Mm. Liam McCabe: If so, I'll do a lookup of the design system schema, again, another JSON, each component has its own type, for example, but a description. And in that description, it's basically saying when is best to use that component. And then from that, it can... generally generate quite a decent page, we find. Although you notice if you use it quite a lot, the pages become quite similar and generic. But then I guess most websites will be, if you think about it. Like you wouldn't want to... Corey Moen: Totally, totally. And like Matthew: Yeah. Corey Moen: the average user's not gonna probably just sit there and keep bashing it. It's us curious devs that do, right? Liam McCabe: night. Matthew: Yeah, let's see how many variations we can get this to sp to kick out. Yeah. Liam McCabe: Hey. Corey Moen: Yeah, yeah, yeah. it looks so good though. Like it looks very polished considering like how you're describing that. Yeah. Matthew: Yeah, we want to show it off now that we've kind of explained it a bit. Liam McCabe: Yeah, so one of the things I was actually... proud of is how we actually decided to integrate it into the page. So this is Corey Moen: Yeah. Liam McCabe: this little ask button. Because when we were using it, we were realizing like this is actually a way to extract knowledge out of our website, or answers, if you will. So the alternative to that is just doing that yourself and browsing. So like any other website with these text links. we're like, well, if you're not going to do that, why even bother having those text links? Corey Moen: Mm. Liam McCabe: It's kind of like an alternative to search, we thought. So basically, when you press this ask button, actually replaces that text nav with the prompt box as well as some suggestions. I think, yeah, I think we're quite proud of that implementation. Matthew: and I see yeah, the so the ask button that kind of opened up the search box input for for asking now says browse. So it's like a toggle between like the two different ways to navigate the site. Liam McCabe: Yeah, because at the end of the day it's still an experiment and we don't obviously Matthew: Mm-hmm. Liam McCabe: want to put people off. We're not going to go to this mode entirely. That'd be kind of ridiculous. Matthew: Yeah. Liam McCabe: But for those people interested, yeah, we kept it that way. And then just having the typical suggestions is also quite nice because some people at the end of the day do just want to click or they do want to test things. And you never know, we actually might get the question right. Matthew: And are those like connected to the top services you provide? Those four suggestions, or how did you choose those? Liam McCabe: So they're random from a long list of anticipated questions we'd think people would want to ask, or actually tailored more to our ICP clients type thing. Ideal candidate. But yeah, I think there's about 20 or so, 30 that it just cycles through each time you toggle. And then, yeah, when you actually... Click on one, we'll ask a question, it will take that query, it will get our design system schema ready, it will get the knowledge base ready, send it all off to an AI model. And then that is, the response is essentially streamed as JSON. So that's why it doesn't just wait for the whole page to load, for example, it will load a component at a time. And that's simply Matthew: Mm-hmm. Liam McCabe: because it's... streaming the JSON, right? So it can detect the type of component it is. So it will say that in the preloader, it's like loading a problem component or a list component. And by the time it's got to the end of the bracket, then it can actually render the component. And then, it's essentially, I don't think it's got all of our components currently, but it's got most of them. Matthew: Yeah. Liam McCabe: And yeah, they're interactive. And then, yeah, we just added a feedback thing here, just in case. Matthew: And does it always end with that kind of contact form? Liam McCabe: Typically, because Matthew: It seems to in my test, but Liam McCabe: I think that is the goal at the end of the day for our Matthew: Yeah, sure. Liam McCabe: main website. Matthew: Yeah. Liam McCabe: Ideally we'd have different CTAs. I think we've only got one at the moment. Matthew: Yeah, I think the I think maybe I've seen like the the heading there in the contact form will like be contextual to what the ask was. Mm-hmm, which is a nice touch. Liam McCabe: Yep, yeah, yeah, that's true. We'll try and tailor this content here, for example. But yeah, I think it really opens it up to A-B testing, if you want to call it that, but I guess it's probably A to Z testing. It's like testing all options almost. Matthew: Yeah, A to infinity. Yeah. Corey Moen: Mm-hmm. Liam McCabe: which could be good, be terrible. But yeah, it's been a fun little experiment for us. Corey Moen: Have you have you know I I especially and maybe this is relevant, I'm not sure. You said that y'all have been working on, you know, AI tooling for clients as well. and something I've heard in the past that can be a concern is like when you're running a model against some, you know, open page query like the fear of spam and racking up API costs and all that. Like, have you had any issues with that where like, you know, it Yeah. Matthew: Are the bots having a good time playing with this? Liam McCabe: Mm. Corey Moen: Have you had any like spam issues or, you know, unexpected hundred dollar, you know, API Matthew: Ha ha ha. Corey Moen: bills for whatever model you're using behind this? Liam McCabe: No, so we have built-in rate limiting and all that. So we've got limits. We've got limits Corey Moen: Cool. Nice, nice, nice. Love it. Liam McCabe: per IP per day and then we have like a cost monetary limit. So should... Matthew: Hmm. It's almost like monitoring your ad spend kind of just its model spin. Yeah. Liam McCabe: Yep. Yep. Corey Moen: Yeah. Like it and I I think that's so smart. And like again, the thing that people maybe often overlook is like there's ways to control these things. and clients are sometimes maybe struck with fear of like, well, I don't know, you know. Liam McCabe: the biggest optimization is caching, right? So all of these Corey Moen: Mm. Liam McCabe: pages are cached. Yeah, so if I was to ask the same question again, whatever it was, it should be somewhat instant. I mean, it's not there, but yeah, it will just serve the cached value in the key value store. Because Corey Moen: Mm. Liam McCabe: otherwise, yeah, it costs like 50 cents each query. Corey Moen: Mm. Wow, wow, wow. Matthew: Hm. So you're doing server side caching of those? Or you're talking about client side? Okay. That's smart. Liam McCabe: Yeah, server-side, yeah. Yeah, it's in a key value store. It's hosted in Webflow Cloud, Corey Moen: Shout out. the other Matthew: Shower up the cloud. Corey Moen: the other thing I wanted to bring up that I, you know, you mentioned earlier that I think is I've also been thinking a lot about is, I mean, yes, the like, you know, huge content corpus idea or whatever, but I think the other part that you mentioned was like you this is another way to understand what users are asking asking. If you're storing these Matthew: Mm-hmm. Mm-hmm. Corey Moen: questions. And what it makes me think way back to is Edgar Allen, you know, they spun off a few different projects in the past. Slater I loved Slater for a time, and then you know, a lot of the model products started getting really good. And then they had one that I think it's called AEO.press. it was that this idea of this, like not the full site design like you've done here, which I think is a much more, you know, user-friendly and beautiful way to show things. Theirs was more like you add a chat bot to your site. so it kind of runs regardless of your actual marketing pages. And people ask questions and then it tries to respond based on your your knowledge base that is just always growing. And then the key is from the AEO perspective is that those questions go back in, and then either the model and then a human has to like review them, you know, confirms what a good answer would be to those. And then it's building Matthew: Mm-hmm. Corey Moen: these like separate static pages for Liam McCabe: Hmm. Corey Moen: it. which The separate static pages are like just text. Like I think the whole strategy or theory was like, will these then get indexed and picked up by agents and be a way to grow your AEO? I have no idea, to be honest, how this is going for them, but I think regardless of that specific implementation, I think this general idea of how do we find more ways to capture specifically what people are looking for on our sites and then Matthew: Mm-hmm. Corey Moen: cycle that back into all of the content that is then feeding the site. But I think I like what you've done here so much more in this realm of like the output, right? Like you're you're feeding back into either generate in real time or over time, you know, actual branded design system aligned pages that are especially like maybe maybe meeting some threshold of if we've hit X amount of questions in this general category, you know, how do we then start meeting that with a more formal, permanently indexed page? You know what I mean? And like Matthew: Mm-hmm. Liam McCabe: Hmm. Corey Moen: especially like maybe more so for like. Clients and marketing sites, right? That like are, you know, agency sites are also growth engine, like for leads and all that. But even for like user adoption, and when we think about like strategies, you know, that I think are still relevant, like programmatic SEO and some of that, it's it's kind of all in that same realm, right? But I think there's something really interesting about how you can create that, like, dare I say, flywheel of, you know, Matthew: Mm. Corey Moen: of effect on the site where you're getting questions, feeding them back in, generating content and kind of like doing that full loop around. anyways, I don't know, just yeah. Liam McCabe: What are your guys view on visiting a webpage and it's not like maybe you've talked about it and it's not the same for each user. Corey Moen: Mm. Matthew: We have talked about this. Yeah, w mixed feelings. Liam McCabe: Cause that seems to be quite contentious. It's Corey Moen: I Liam McCabe: like when all the feeds right these days in X and Reddit you go on it and it's just never the same. can never find what you were just looking for. Corey Moen: Totally. I mean, that's especially if you are truly, like you said, searching for something like even an X, I swear I'm like just looking for somebody that I just saw and I'm like, why isn't it coming up? Like, Liam McCabe: D-doesn't help that the search is also crap. Matthew: Or yeah, you you see something, you're like, that's interesting, you click or swipe accidentally and you try to go back and it's like it's gone. How do what Corey Moen: Yeah, yeah, yeah. So Liam McCabe: Yep. Matthew: how do I get back to it? Where do I find it? Corey Moen: my my wonder here in all of this is like personalization is already a thing kind of on the web, right? Like even Webflow Optimize offers some personalization functionality, or you know, there's more and more products coming out that claim that they can like de-anonymize users to an extent, right? Like it of course get their location or maybe their company or like in a creepy way, maybe their email to try to personalize. But I think. Like my wonder is, will browsers ever get to the point of social media, right? Where they're able to know more or people are willing to opt in more of their personal information or what they're doing when they browse, in order to get a more personalized experience? And would that become the norm as it has become the norm on social media, which to me is ironic, right? Like the there in browser land, it feels like at the moment it's almost the opposite. There's like privacy first browsers coming up out of the woodwork every single day because people are like, I don't want to share any of my stuff. And then those exact same people are opening Instagram or TikTok or whatever or YouTube for Liam McCabe: Yeah. Corey Moen: me. Like, and it's just like all all of your stuff is recommendation based, right? so I don't know. Yeah, Liam McCabe: Yep. Corey Moen: I think it's a great question to pose. Like, I mean, Liam, do you have any takes on this? Liam McCabe: I yeah, I tend to think maybe even too far down the line. Like, where will websites be in 10 years? Like, especially with all the AI services that are coming in and they're playing around with generative UI, right? Corey Moen: Mm. Liam McCabe: If you can get your answer quicker than browsing a website, you just won't go to the website, right? I'm sure there's lots of reasons to actually visit the website as well for visuals and actually it's much more attractive Corey Moen: Mm-hmm. Liam McCabe: way of digesting the information. But I did wonder if we're heading down like a, I think it's Google Stitch route. Did you ever hear of Google Stitch, which was an experiment from Google where it was back in the day, it used to be a browser and it was basically a prompt based browser. or query based browser and it would take your prompt and basically generate the webpage, kind of what we're doing for Code and Wonder, but on a browser level. So Corey Moen: Right. Mm-hmm. Liam McCabe: for any type of website brand company. And I'm wondering if an AI service would go down that route, but it feels like they've tried building browsers and they've just like given up. I guess it would require like a standard for design systems Corey Moen: Mm-hmm. Liam McCabe: that knowledge base. Like if every company had that, then you could just put an Aquarium and it would return a branded response. Corey Moen: Mm-hmm. Liam McCabe: yeah, the amount of collaboration on that and yeah, it'd probably be a bit too difficult. Corey Moen: For sure. I think it's Liam McCabe: too much of a challenge. Corey Moen: I think the the like company knowledge base thing though is like there's more and more people talking about this. And is it a matter of time before there is some solution? Like we just posted on No Code Supply Co. It kind of went viral. There was an article from the head of the brand at Ramp. Did you see this? The brand as software Matthew: Mm-hmm. Corey Moen: article. If anybody hasn't, it's like at near the top on the No Code Supply Co homepage. And it was it it still ladders up to this exact same primitive of like basically saying, you know, that teams, mainly in this case, in-house teams. But again, I think I would argue this applies to agencies, especially the agencies that I've worked at in the past where you do have, you know, multiple designers and devs and whatever is that like for ongoing work in the agency case, retainers and that kind of stuff in in-house, right? Like all of your stakeholders, you have a queue. And that queue is just never ending of requests and random stuff. Matthew: Mm-hmm. Corey Moen: And and Liam McCabe: Hmm. Corey Moen: AI in many ways is almost making that like worse because your clients or your teammates or whatever is now moving faster or producing more and then you Matthew: Mm-hmm. Corey Moen: have to react more. And like it's just I think the main point was that it's kind of just a workflow of the past that doesn't is maybe not as relevant anymore, or we need to rethink how we work as teams in this age. And and where all this ladders up to is then this idea of brand as software, or again, like a you know. A huge knowledge base of everything you would ever say about your company or your brand that is always growing. But the main point is that it doesn't necessarily have to be like, like you were kind of mentioning, like it doesn't necessarily have to be like every Slack message and like everything anybody's ever said. It's actually just like, what's all the final output? Like, what's all the like, dare I say, approved, you know, copy about a brand or a company that is in this like huge knowledge base? And the main emphasis being that like. To your point, Liam, like is that what will actually become more valuable in the future? And it's less about like how you put it out on this website or that email or that social post or whatever. The main goal is just to have this like really high quality corpus of all of the things you would ever say. And then as the industry changes, as the outputs change, maybe people don't want websites as much anymore. We're sending this knowledge directly to agents some other way. I truly have no idea. whatever Liam McCabe: Hmm. Corey Moen: it is, the point is is that you'll be set up to be agile. You'll be set up to know like these are the things we wanna say however we wanna say it, you know, and we'll keep working on it that way. I don't know. Liam McCabe: Yep. Corey Moen: End rant. It's just Matthew: The the brand corpus is gonna get so big that there's gonna be this new role invented called a brand librarian. The people who are actually charged with maintaining making sure everything is on Corey Moen: I love it. Matthew: the shelves as it should be in the right place. going back to browsers and personalization, I'm kind of having this thought like the browser seems optimal to kind of house your personal preferences. And be able to have some kind of API between the site you're visiting, the brand that's serving up the content, in a way that is anonymized, but Corey Moen: Mm. Matthew: still allows you to serve the personalized content. But I think that requires us trusting the people who make the browsers and obviously with the rise of privacy first browsers, that is a is a real big issue. Corey Moen: Yes, risk for sure. Matthew: So so yeah, other otherwise it's like I mean that's why we have these tracking pixels and you know cookies and all this kind these those things were invented because brands were frustrated that they didn't have the controls to personalize content, right? So that's why Facebook, you know, created the pixel and you know all that. And Liam McCabe: Hmm. Matthew: you know, that's why we have these creepy patterns on the web. It was because The companies that could kind of control and give us like a safe way to do that in a privacy first way just were too slow. Liam McCabe: Yeah, mean, even if you just take the non evil route, right, you just you just want to deliver the content that the person is looking for. Right. Matthew: Mm-hmm. Liam McCabe: And that does require a certain amount of information. Some of it might be personal, some of it not. But ultimately, you're trying to just reduce that amount of searching, that journey to get the result you want, I guess. Matthew: Yep. We all want to get to the destination as quick as possible. Yeah. Liam McCabe: But also, but also it's about the journey, right? You got to enjoy the journey. So it's not just about the destination. Corey Moen: Facts, facts. Matthew: Sorry. Corey Moen: I Liam McCabe: You gotta browse these lovely fancy websites and enjoy it. Matthew: That's right. Corey Moen: This Matthew: How else are they gonna get awards? Liam McCabe: Yeah. Corey Moen: think I think it's interesting too, because there's definitely the the the personalized website thing for me is also like, I don't think we necessarily have to wait to even to what you're doing here, right? On the Code and Wander site. Like we don't have to wait for browsers to eventually do some of this stuff. Like or even to the point and and I'm not saying that from the lens of like just be nefarious and start scraping data and Matthew: She's Corey Moen: like, you know, breaking privacy things that way. I'm more meaning like could C is there other ways that we can simply serve up at least like major persona versions of our site or pages, you know, through direct things like a URL parameter or something b from an ad or from an outbound email campaign to people that you know are marketers or executives or engineers or whatever. Right? Like, yeah, I don't know. It's it it feels like there's a lot of opportunities there. Matthew: That's true. Corey Moen: Yeah. Matthew: Like personalized campaigns based on like personas or some kind of bigger group, not so necessarily individual as far as personalization. These are methods we've had available to us, but maybe we haven't leaned into them enough. Corey Moen: Totally, totally. Yeah, like even dare I shout out webflow.com, it it was rebuilt last year and it has this like basic premise of this where like in the nav, there is here I could share my screen Matthew: Yeah, Corey Moen: yeah, even even on on webflow.com, which was rebuilt like a year ago, they have this this premise of a few places in the site where you can like do very basic stuff of like, yeah, sure, I can switch content just in this section, but that actually like follows you around the site. So you can see, like, here if I switch to designer, it switch here switches here. And the whole general point was that like you could enter this way and set this cookie via URL param. So that way, like outbound campaigns and all that stuff that's sent to any existing landing page on the site could already have some indicator of what they think that audience is. And then the theory was then that way when you go around the site and go to different pages, I don't even know what's what here if they're still using this much anymore. But the thought was that then that would already be set to that persona for at least a few sections. And like What's funny is the original vision for all this was much grander. It was like, we want this, like, you know, in we don't even need the URL prem outbound thing. Like, can we just detect these things on the fly with good webflow optimized? And like, you can do some things, but again, not granular to a persona. It's like it's like I said, at least like through, you know, segment or whatever those those out outside toolings are. I guess there is the toggle here in the hero. Matthew: But it's like the ask button Liam McCabe: Yeah. Matthew: that Liam built. You know, it's like whenever the visitor can actually declare what they're looking for or what they're interested in, that's a lot stronger signal than trying to intuit it from sources. Corey Moen: Totally, totally. Liam McCabe: Yeah, think I'd want it on the browser level. I'd want to see a little pop-up in the right that says, this page has personalization options. Do you want to apply them? And so basically Corey Moen: love that. Yes. Liam McCabe: the browser can hold any sort of personal information that you want to put into it. Like, I'm from London, Corey Moen: That you Yeah. Liam McCabe: I'm an agency founder. And then either you can set it to always be on or... off or it can just be in between and on per site per approval or something like that and then yeah just it removes all those toggles that you see on this landing page Corey Moen: For sure. And and when you think about we have basic stuff, like we can detect, you know, light dark preferences through your OS in the browser. Like there is like very simple things like that today, a reduced motion, whatever, right? Liam McCabe: Hmm. Corey Moen: So I a hundred percent agree. Like that would be so nice if browsers would adopt that ability for at least users to be willing to opt into that. Because I even from like a simple lens of things like shopping online, right? Like I would love it if I could just and maybe this exists and I just and don't know. Could I just like put here's the shirt size I usually wear, pants, shoes, and again, my general preferences for aesthetics and that is just held in my browser on mobile and desktop and a so associated with profile. Yeah. Matthew: So every time you visit a shopping site it already and you visit a page is already like preloaded with the selected size that you Corey Moen: Right, right. Like, and again, I realize that individual sites store that, like they Liam McCabe: Yep. Corey Moen: are intelligent enough in that way, but like why can't we do this at the ecosystem level, at the browser level, like you're saying, Liam? Like I think it just makes so much sense and like I don't know, people jump so quickly again to this like, I don't, you know, I don't want to share that information. It's like you're already sharing that information whether you realize it or not, all over social media, like Liam McCabe: I mean, yeah, the problem is it'd be opt-in, right? So majority of people probably Corey Moen: For sure. Liam McCabe: wouldn't opt-in, so companies would be throwing away a bunch of data. Which they're never going Corey Moen: Yeah. Liam McCabe: to do, right? Matthew: Unless you put a a sight gate like, you know, for sites with alcohol and you have to say that yes, I'm over eighteen or whatever. I I do remember as far as like Corey Moen: Yeah, yeah, yeah. Liam McCabe: Yeah. Matthew: privacy concerns with these things, I do remember back to and vaguely remember so I can't remember the specifics, but when there were debates in like whether or not I don't remember if it was at the browser level, the computer level, if it should remember if you prefer light or dark mode and stuff. In the accessibility community, there were concerns that pe they would start being tracked and identified because of, you know, having those accessibility preferences being passed like through into the browser. you know, I think thankfully those concerns I've never heard of those concerns actually being Liam McCabe: What? It's- Matthew: validated and being violated. you Liam McCabe: Yeah. Matthew: know, but I think it we've kind of had these conversations before. It's just each time we're gonna take a l step a little bit further and it's like you know, slowly stepping into the boiling water. Liam McCabe: The problem is though, even with dark mode and light mode, you can put those into buckets, right? The types of people that choose those and Matthew: Mm-hmm. Yeah. Liam McCabe: you can put them into more and more buckets. So it's actually quite revealing Matthew: Yeah. Liam McCabe: when you say what mode you're in. Matthew: Yeah, it can be. And then there are folks like me who have it on auto because I like it dark at night and light during the day and then I just confuse all the data. Liam McCabe: really? Yeah, it's strange. I'm in dark mode a lot of the time with programming and stuff, but I like to see nice designs in light mode. Matthew: So it's like task so for you it's like a task delineation. It's like if you're doing dev work, you want it be dark. So you like you know you're in code mode. Liam McCabe: Yeah, otherwise it's just in your face, right? It's like, I can't do that. Especially at night. Matthew: Yeah, yeah. Corey Moen: So good. I used to be a a big dark only fan, dark mode only fan, and now I don't know why, what changed. I am I'm the opposite now. I do light in pretty much everything. Liam McCabe: Yeah. Corey Moen: I maybe it is I just I definitely look at code far less than ever in my career. Matthew: Mm-hmm. Corey Moen: and so I don't know if it's that of just like reading in the different context or I don't know. maybe Liam McCabe: Mm. Corey Moen: it's just a season. Matthew: I mean books were never like black pages with white text, you know, they don't sell books like that, so Corey Moen: Fair, fair, fair, fair. Fair. Matthew: I wonder why. Liam McCabe: Very true, very true. Corey Moen: Yeah. Yeah. Sweet. Matthew: Liam, I was curious going back to the what I don't the ask functionality on the side. I don't know how you how do you refer to it? The ask agent? I don't know. Liam McCabe: Yeah, we call it Ask Kodamwonder, I think. Ask Kodamwonder. Matthew: Yeah. so for that functionality, are are you do you have any kind of you said you're caching, you know, results and stuff. So you have some kind of insights that you can get out of what people are Liam McCabe: Hmm. Matthew: querying. You know, are you I mean, maybe it's still too early, but are you looking at, you know, how can like Corey mentioned, like we create new content or maybe it informs like this should actually be its own, like Dedicated landing page. You know, is it is it feeding back into the design of your site or do you have plans for that? Liam McCabe: Yeah, think so. guess similar to like just search functionality, it's interesting to see what people query and it has basically raised a few content aspects that we're not actually including on our website, which we probably should. So it has actually been quite revealing in that sense. It's like, yeah, even something as simple as Arlian and Alessia, the co-founders site. I don't know. It's not on the website. I'm like... What? It should be on a website. mean, because Matthew: It's not on the website. Right. Liam McCabe: we're directors in the about page or something like that, but it's like, yeah, that should be kind of obvious. So Matthew: Mm. Liam McCabe: that's a small copy change that we can update on our website. that there has been a few instances like that. Matthew: So that means kind of as individuals, y'all have visibility in different places online and people are trying to make the connections between Code and Wander and and yourselves. Yeah. Liam McCabe: Yep. Yep. yeah, in that sense, it has allowed us to basically update the website in a more for the audience way, I guess, rather than guessing Matthew: Mm-hmm. Liam McCabe: a lot of stuff, which is incredibly valuable. In the long run, I think we want to obviously try and measure conversions and which... which pages are performing better and which pages compared to the static versions are performing better. It's still early days at the moment, but once we have a bit more traffic over the past few months, that'll be an interesting analysis to do, I think. Matthew: Yeah, I think so. Cause like if we just look traditionally, what have we had to work on as far as site analytics? You know, it's it's it's clicks, it's loads, it's it's time on page, you know, and none of those explicitly tell you what the visitor is looking for, searching for, you know, what is it they want Corey Moen: Yeah. Liam McCabe: Mm. Matthew: to see? But by building this ask functionality, you're actually able to start getting you know, site into You know, why why are you why are you really here? What are you actually interested in learning? You know, and then we can actually start having a conversation rather than the site just trying to be like, here's everything we think you might Corey Moen: Mm-hmm. Liam McCabe: Yep. Matthew: want to know. Liam McCabe: And I think that probably wouldn't have been possible prior to all the AI services, right? And the Matthew: Mm-hmm. Liam McCabe: conversational view, because Google essentially taught people not to ask questions, but search for keywords and use various ASCII and stuff. But now we're back to these Matthew: And now all we do is start with questions, right? It's question first. Liam McCabe: long form questions. And yeah, I think those... Yeah, again, are a bit more revealing. It's like, they're actually looking for this, rather than if you search for that keywords, that doesn't mean they're actually wanting to know about that. It's just might be related. So yeah, it's really interesting just seeing who clicks on the buttons and who actually writes a query, what those queries are about. A lot of them are about services, like, do you do this? Do you do that? Because sometimes I guess, you might have something in your knowledge base that you don't necessarily want to put on the website, or it might not be indexable, or it might be hidden in some way, but you still want people to Matthew: Hm. Liam McCabe: know about you. Corey Moen: Mm. Liam McCabe: And that can be a way in a search input, or in fact now a prompt box. So it has been a way that we can supply additional information without crowding the website out, you know? Corey Moen: Interesting. Have you has the queries that you have observed like surfaced any services or things that you haven't done in the past that has brought up new conversation internally? Like Exactly. Yeah. Matthew: Like you're like, there's demand for this. Maybe we should serve do it. Liam McCabe: Maybe we should pivot and start making furniture. Matthew: Well, it's like, you know, who who is going in here and asking for where do I find a Diet Coke close to me, you know? Liam McCabe: Yeah, certainly, there's a few services that we've never really done like e-commerce and people are asking about that. It's like, maybe we should finally jump on e-commerce, I can... Obviously Shopify is booming. But yeah, it does just provoke those conversations, right? And every week we get to have those conversations and it's like... maybe we can redirect Corey Moen: Super cool. Liam McCabe: towards this direction or towards the other direction. Matthew: Yeah, see where the demand is heading. Corey Moen: Yeah. And like even to what Liam McCabe: Yeah, but, go on. Corey Moen: you said, it the more traffic you get, the more you could almost maybe even see trends, right? Like it just the the thing that popped in my head was like, you know, five, six years ago, like especially with Webflow, it felt like a lot of agencies and freelancers I talked to, they'd you know, they were like selling Webflow. Like people would be like, I need a website, and then they'd you'd be convincing them of Webflow. And then it shifted to like more people were just getting requests on like, I know I want to use Webflow. you know, and then they're just trying to find the right person. So the point is if even this kind of functionality existed then, would you start to be able to see those trends of like maybe now it would be like Rive or something, like somebody's looking for a very specific type of animation tooling or something like that that they already know they want to use, they're looking for the right partner. It's like that aspect of this I think is really interesting to be able to observe. Liam McCabe: It was, yeah. It's also a potential teaching moment as well. That's right. Like, Corey Moen: Hmm. Liam McCabe: we don't, we don't offer Rive, but we also do offer these two alternatives, which can produce a similar result. So. Corey Moen: Totally, totally. Right, right, right. Super interesting. Liam McCabe: Yeah. Matthew: Cool. Do you have any ideas or plans to iterate on this? Liam McCabe: I have left it for a while, but the last thing I was working on regarding this was when you arrive on a generated page, does that have to be the end of the funnel? Where Corey Moen: Mm Liam McCabe: do you go from there if you're not filling in the form? The idea was how to keep them... keep them asking basically because they can just ask another question in the prompt box. But I kind of wanted people to be able to do it anywhere. Matthew: But that's almost like a separate separate journey if they do a new input, right? Not a continuation. Liam McCabe: Yeah, so it would be basically like what would be the equivalent of clicking on a link on a page Matthew: Mm-hmm. Liam McCabe: in this AI world. So the initial idea I had was just some sort of annotation. So you could just annotate, maybe not like highlight. Currently it's highlighting, but maybe it's just click. can click anywhere. And basically a little prompt window comes up and that is a bit more intercom style. And you can say, I don't understand this. What is this? Or ask more details about this. And currently that's actually hooked up to our Slack. Well, it's all hooked up to our Slack. So we could then jump in and reply to that query and it will reply in line. But there's also that idea that it could just generate another page. So you could click through to another page or... Matthew: Or even just a comment, right? Like if they're selecting or clicking something that's in one of those generated sections, it can just, you know, allow them to ask a follow up question and then you could Liam McCabe: Yep. Matthew: just answer it in line. It could be lower Liam McCabe: Yep. Matthew: fidelity than having generate more UI. Corey Moen: Yeah. Liam McCabe: Yeah, exactly. That's what it currently does. yeah, because it's hooked up to our Slack, we've also hooked it up to the knowledge base as well. if the answer is unsure, or the AI is unsure about the answer, we can kind of jump in on Slack and we can say, we actually do this. So then it will update the knowledge base. So the next time around, the AI should have the answer. but yeah, as, as, as it also appears in like a little intercom box, we can also reply in Slack and it will show on the website. but yeah, I'm, yeah, I'm unsure about that whole experience. I don't think people are used to just clicking around, annotating websites. Really. It's not a, maybe in our field, maybe, maybe, Matthew: Yeah. How discoverable is that? Yeah. Liam McCabe: maybe in our field slightly, but the majority, I don't know, people are just used to. I know some people highlight things to read them, guess, but most people just click around and scroll, right? There's Corey Moen: Mm. Liam McCabe: not much to it. Corey Moen: I think that is the part, even what you said earlier about like Google has trained us to search with keywords and things like that. Like, has the has the internet and the browser, sorry, this is kind of meta, like existed long enough. And at least for the foreseeable near future, it's like that browsing experience, especially when it comes to shopping for anything, which includes services, right? Like who to work with on something or whatever, like that browsing experience I think is still Valuable to an extent. It shows some level of like trust building or whatever, like you can get a sense of quality, just like you can get a sense of quality if you go into a store and can touch and feel you know the clothes or whatever the product is versus just seeing it online. Like, you know, I I don't know, like cause it so many of these things always come back to like will we just do we even need websites and will we just, you know, all ask agents to self is like I I don't know, or or even the I've seen, you know, some of the model companies doing the like render, like, doesn't ChatGPT do this where it'll like render components into the chat or like, you know, aspects of a Matthew: Mm-hmm. Corey Moen: brand in so that you don't need to go to the website. And it's like, again, though, I don't I wonder how that's going. I to be honest, because I'm like, I maybe it's just me. I still want to go out and see the real thing somebody made, which in this case for us is websites, right? Like, I don't know. Liam McCabe: I think there's merits to the experience, And Corey Moen: Right. Liam McCabe: actually immersing yourself in the brand. People that Matthew: Mm-hmm. Liam McCabe: buy things often, or a brand that they like, they tend to want to immerse themselves in that brand and learn as much as they can Matthew: Mm-hmm. Liam McCabe: about them. Maybe not as much, but some people do. You get obsessive fans and stuff. yeah. Matthew: That's why they call lifestyle brands, right? Mm-hmm. Corey Moen: for sure. For sure. Liam McCabe: So it's like you'll get as much as you can from the experience and that experience is only going to build your brand even more and your reputation even more and they'll want to come back and back. Whereas if they serve the result immediately it's like, that's done what do I do now? It's like that wasn't as fulfilling maybe than the Matthew: Mm-hmm. Liam McCabe: experience could be. Playing devil's advocate here rather than a doomer. Corey Moen: Mm-hmm. Matthew: I I agree. I think as humans we like destinations and experiences. And we've seen a growth in those types of industries in the past couple of years with with the rise of AI and you know the ability to just do everything in a isolated, you know, chat stream. Corey Moen: Mm-hmm. Matthew: We've actually seen a rise in in person events and con concerts, you know, and you know, comedians, whatever it may be, all these kind of in-person experiences and tra destination travel is, you know, bigger than it's ever been. Cause that's what people like. That's you want you wanna feel that immersion, right, in into Corey Moen: Yeah. Matthew: the thing. And one thing that comes to mind from the brand perspective is you You should not want to give up your website or or give up on it because that is the thing you own. Corey Moen: Mm. Matthew: Right. That is the place of the internet. It's your street address. It's like having a physical shop, but digital, right? You know, where people can go and they know they can find you. And it's your one chance to tell your story. Whereas on all the other platforms, whether it's inserted dynamically into an AI experience, you know, or in social media platforms. it's out of your control. You can kind of create something and put it on there, but who it reaches, you know, how and when they see it, you know, it's all out of your hands. so I I think we we've Liam McCabe: Yeah, agree. Matthew: got to hold on to websites because it is it is kind of real estate that you own versus, you know, real estate that you lease on platforms that that are owned by Liam McCabe: Yeah. Matthew: big comp big companies. Corey Moen: For sure. Liam McCabe: It just seems like an element of freedom. Yeah, we need this stuff. We can't Corey Moen: Ha ha. Liam McCabe: be tied to just five tech companies. It needs to be Matthew: Mm. Corey Moen: my goodness, this. Liam McCabe: the wild, wild rest. Matthew: Yeah. Corey Moen: Yes. Matthew: Yep. Well Liam, is there anything you want to ask us. Liam McCabe: Let me think, let me think. I'm interested to hear your thoughts about source as much as you can say. Matthew: yeah. Liam McCabe: I mean, not just because it's Webflow, but I'm very curious because they seem to be, yeah, I don't know how much you want to include, but they seem to be going more corporatey. And I understand that because a lot of websites will be becoming more like... simple design system generations, but I do miss the fact that they're not focusing on the creative aspects of the designer and that sort of stuff like that. I guess, do you see a rise in more creative, quote unquote awards type websites, or are we going to see more of these design system generation type websites and we'll be focusing more on growth engines, etc.? I guess this is Speaking more about marketing websites, I guess. But yeah, I guess I've been thinking about that. Like if websites are gonna stick around, I think that obviously they will stick around in the near future. But what will be the breakdown, the types of websites that will? Obviously we're still gonna have the big community-based websites, the regulars that we check in every now and then. But what will marketing websites evolve into? if anything. Matthew: I mean, I've I've always felt this about websites. building like brochure type websites where it's like this could have been a PDF or this could have been an email websites never excited me. It was always can I create an experience or can I tell a story you know, about about a brand or about a product? That was always what interested me in building sites. And I think with all the things we've been talking about of the trends in in marketing, AI and stuff, I think those type of websites are gonna be the ones that stand out and the ones that people talk about. I don't know if all websites will kind of fit in those categories of experiences. But but I think those as being the strongest ones. And does that that doesn't necessarily mean the award space sites that are like just highly animated and interactive. it can those fit into that category, but that it doesn't mean it's only those kind of high end visual experiences. Liam McCabe: So, AI services where they are currently won't replace any more. You think they won't focus on more UI generation? Matthew: Yeah, I think UI generation is definitely going to become prominent over the next few years, you know, w whether that's you're generated on on your site or within other platforms. So I think that is a big, big reason it's important for people to focus on those brand corpuses, knowledge bases, those things and start being able to stand those up and have them ready because yeah, the time is coming. when you're gonna need those, you know, whether it's you're generating UI in your your app customer facing app, your website, or within other other platforms. Corey Moen: Mm-hmm. Liam McCabe: Mm. Corey Moen: I think we're also in a time too where there it is, it is we can make a lot of speculations and we can try to predict what's coming and anticipate and plan for it. But like I just keep going back to that that realm of like, how do I just work in a way that I can be as agile and ready for however? the web evolves and changes, right? And in the context of even like source, and some of that, and like even some of the articles that came out ahead of conf, I think was interesting and like getting back to code and all of that. Like, I just that to me fits right in that bucket of like if we can just work with a tool, whatever it is, that is back to just open web frameworks, you know, and I mean, you know, a everybody's talking about Astro and all that in the Webflow community these days. But like Astro's been around for for a bit. Like I was joking with Matthew the other day, like Ben Parker is somebody that that is just a legend at Webflow. And I r I visibly remember like, I don't know, two, three years ago, I think it was, that he was like, Have you checked out Astro? Like this is like right when it came out or something. And I was like, I just I again I just kind of like blinders on. I was like, I don't want to go back into an IDE and all that. Like I'll just stay in my comfy little web flow designer. And then now that, you know, we have agents or whatever that can help us stand up some of that stuff. And now I'm just like, why didn't I look at this sooner? Like there's so many things about it that are just like respecting HTML, CSS, and JavaScript in their vanilla forms that I have always loved. And I didn't ever I was more hesitant to go down the like holes of frameworks. Like, you know, Just in the same way that jQuery was once amazing and now is completely dead. Like I I always had this underlying fear of like, well, will the same thing happen to React or next or whatever? I don't know. Today at this moment in time, people are like, never. It's like, I'm sure people were saying that eight, 10 years ago at jQuery, too. and so the point is though, if we're getting back to code, if we look at those as examples, it doesn't matter. Like a new framework could come out. And you can just shift to that. But you know what doesn't change is your words, is what you say about yourself and your company. And like that, those things are evergreen. And so, like, if everything else can be agile, then like how do we work on that to prepare for whatever it is that comes, you know? I don't know. And so, like, even sorry, I'm saying all that in the context that like you'd asked about Webflow Source. And I Matthew's like, I don't work there anymore, so I'll happily talk about it as like a community member. It's like Matthew: No, say all you want. Go for it. Corey Moen: The one thing that I think, you know, came across about source at Webflow was this like, and then in our last episode, Ollie mentioned it very well this way of like the like cursor for marketers. Like there was in general a lot of focus on people feeling like this is all for marketers. But I think the the thing that was missed is like, y'all, for us as developers, like the point is it's all back into code. So you do whatever the heck you want. Like it it's less about your workflow because now you can do it literally ever any workflow you want instead of webflow defining your workflow. And now Webflow is just trying to shift the priority to how the marketers and people that you serve ultimately as your clients or your team, how they work and how you can enable them with that. Right. So like there's a part of me that's like rad, like I'm just gonna get stuff in Astro or whatever and I'll just build and do what I want. And you know, we we spent enormous amounts of time trying to perfect our workflows and web flow to make it easier for clients to edit. But like let's be super real. Like, is there any easier way for them to make an edit or a change than just asking a agent to do it? Especially like at this moment in time where the agents are getting pretty darn good at at interpreting that, right? Like some people still think about them like a year ago or more, where like they would get it wrong half the time. And to be clear, they're not perfect. But like if we look at that lens of what's changed in the last year, like it's gotten pretty, pretty good, right? And and there's a lot of ways we can put those guardrails in place. And and at the same time, I'll say that like I've just always had this perspective of like, it's I I very much empathize with the fear of like, well, if they can ask the client the agent to do that, what am I here for? And it's like, that is absolutely a valid way to look at it. I just I I just think there's way more exciting things we can be doing than making spelling you know fixes and Matthew: Mm-hmm. Corey Moen: fixing broken links and swapping an image and like just junk that we spend, let's be super real, like over 50% of our time doing on an in-house team or even for agencies that do work with retainers or in that last mile of QA, right? Like, so what if we can save all of that little tedious stuff and then have even more time and energy to put into like the really impactful. Things, whether it is storytelling or growth flywheels like we were talking about earlier, or things like Matthew: Yeah. Corey Moen: that. You know what I mean? And so, like, that's the focus of source that I think is missed. and and I it'll just be really interesting to see like it actually implemented and how people start using it. because like I think there's still a lot to be wondered, like pricing and services, and like why would I use this over just like Claude Desktop or cursor or codex or any of these other options? Like, Genuinely, I don't know, but I do arguably think it is at least more in the realm of where things look like they could be valuable and going versus closed platforms. Like, you know, even some of these other tools that have come out, like Ploy or you know, even dare I say Framer or Lovable, those are still like pretty darn close. Yes, they are like closer to code. But like it is not as much you can just use anything from the web ecosystem and start building your site on it when it comes out, versus like if you are just back to the source, pun pun intended, Matthew: Mm-hmm. Corey Moen: you know, like does that open up that freedom? And I think that entire thing was, I mean, if I had feedback for the web flow team, it's like, I think you missed a little bit of that emphasis and you maybe went a little too hard on the marketing talk, which again I get is is valuable, but like. Liam McCabe: Hmm. Corey Moen: Don't over over yeah, overlook the value of just like that that open ecosystem. So anyways, we'll see. Yeah. Liam McCabe: Yeah, no, I'm excited. I'm excited to try it. Because it feels like it has been kind of following our agency model the way we've been going the past two or three years. So it Matthew: Mm. Liam McCabe: kind of slots in nicely right there. And Corey Moen: Love it. Liam McCabe: the fact that it's a bit more flexible with the builds and the whole extensibility. Corey Moen: Mm-hmm. Liam McCabe: I'm curious. I'm very curious. until I get my hands on it. Yeah, we shall see. But Matthew: Not not too not too much longer. Not too much longer. Liam McCabe: I do actually have another question which I've remembered, which is, did you see the Sketch co-founders new teaser thing? Corey Moen: Yes. Matthew: Mm-hmm. Corey Moen: E is it Elix, right? Liam McCabe: Yeah, how do you say it? E- elix, elix- Matthew: yeah, how do you say that? Elix? Corey Moen: Yeah. Matthew: Like el like Elixir? Like Elix? Corey Moen: I I saw somebody mention yes, is it a is it a play on Elixir? I don't know. Yeah. Liam, do you have Liam McCabe: Hmm. Corey Moen: any do you have any takes on the limited amount of information they gave us about it? Liam McCabe: I have no idea. I basically try to think if they're going down more of a publishing route or Corey Moen: Mm-hmm. Liam McCabe: a more of a generation route and I'm undecided. Maybe both. I don't know. Corey Moen: Yes. I Matthew: It yeah, I think it's still fuzzy, right? What this actually is. Corey Moen: Totally. I they I so they did the launch post and then I saw I think yesterday maybe they they had a follow-up of like a little bit Matthew: Mm-hmm. Corey Moen: more context on like talking about components and starting starting granularly and then working your way up in how you build components and then sections and it's like, so like atomic design systems that have been around for 10 years or like there was a lot of, you know, harsh comments on like so you mean React. Liam McCabe: Yep, Corey Moen: like Liam McCabe: yep. Corey Moen: it there's just it doesn't seem like there's much out of it that feels truly novel and unique. But again, maybe that's to be exposed and and maybe like the path I could see them going is if you're building with Elix, however you say it, components and an entire system, does that mean that you can manipulate that both in code or in a canvas? is the one speculative thing that I wonder is where they're going. You know what I mean? Like kind of like what paper's doing, kinda like what Figma is trying to do where you can like render localhost in in the canvas and like how did you see that at like config where you like can dupe a frame and that creates a new branch. I think that was like really compelling. I've never used it, but Matthew: But e the if did you see Elix is actually the file name. So it's actually a Liam McCabe: So yeah. Matthew: custom file type like they're Corey Moen: Mm-hmm. Matthew: doing, which Liam McCabe: Mm. Matthew: like Affinity has done. But it's like I don't know, like there's a lot of debate internally at Webflow with those of us who are like connected to the design world. And yeah, people are not happy about that. It's because it feels like lock in again. Corey Moen: Right. Matthew: And it's like, Corey Moen: Right. Matthew: I thought we were past this. Like we should be using open standards and and code. yeah. Corey Moen: But is the is is the file type, I wonder, in a literal form, like you said, like an affinity? Or is it like, I mean, again, a still learning, getting the hang of Astro, but Matthew: We don't know. Corey Moen: like Astros dot astro files, and isn't that solely for a purpose? It yeah, it's not it's Matthew: But it's just a text file. Yeah. Liam McCabe: It's just JSON. Corey Moen: not a proprietary language. It's just for the purpose of recognizing and build time or something, right? Like i Matthew: Yeah. Corey Moen: is this the same kind of thing that it's you know, it's maybe not lock-in, it's just how it gets expressed. Matthew: I think maybe that's Corey Moen: Yeah. Matthew: that's the best thing we could hope for is that is is is just so that the app can recognize the files it can like open easily and that once Corey Moen: Mm-hmm. Mm-hmm. Matthew: you actually inspect the file and the contents inside that it is maybe it's just JSON, right? Like we're talking Liam McCabe: Mm. Corey Moen: Yeah. Matthew: about everything's JSON. yeah, it'd be interesting. Liam McCabe: I do find a lot of these tools seem to just be converging, you know, onto the same thing. And I do find it slightly concerning that they're doing what all startups shouldn't really do, which is depend. Your main business model depends on another service or company, right? Like everyone's Corey Moen: Yeah. Yeah. Liam McCabe: just relying on the AI companies, which could... could just turn everything off or they could just replicate your whole business. So it's like playing a risky game. Corey Moen: Totally. Matthew: Yeah. Or they just are they just angling to get bought out by one of the companies, you know. Corey Moen: There is that. There's a lot of rappers out there that that is their actual goal, right? like, yeah, or look, we Matthew: Yep. Yeah, just to make enough noise to get yeah, scooped done. Liam McCabe: Yeah. Corey Moen: look at this novel idea we had on how to apply or productize some aspect of models, you know, like 100%. I I think the other thing that comes to mind with this, and I'm curious, Liam, as this even relates to Code and Wander and how your team's workflow has evolved at at, you know, or not, is like. I I you know, Figma, I think for just speaking from my own use case, I think Figma is still so useful for like early ideation, like throwing mood boards together, doing some really, really quick explorations of like layouts and compositions and ideas, but then like moving to code as quickly as we can, and then iterating further from there and actually implementing. And so I'm saying all that in the context of like, Especially a lot of times what we're handing off as an agency or freelancer or even, you know, what we're maintaining as an in-house team is a design system. And I think there's still some like gray areas in the industry at the moment where team some teams are absolutely still maintaining design systems in both Figma and Code, like we've done for the last, I don't know, five, seven, eight years, and and even earlier with sketch, ironically, right? Like, or even before that, dare I say, in the old school Photoshop and Illustrator these days. Matthew: Yep. J I used to just have color swatches in an illustrator file. Yeah. Corey Moen: Yes. Yes. And I'm and where I'm going with all this is I think I'm never I've never been more convinced that like so many of these tools are getting good enough that even designers who don't want to touch code or ask an agent or whatever, that's fine. They should still have a visual way to go help maintain that system, that design system, but it ideally is just tied to the code. Code again, sorry, this one is no pun intended. Code is the source of truth. Let's be super real. Like you know, and so it's kind of mad that teens still have two sources of truth and have to keep them aligned in Figma and in code and they drift. They will always drift. That is inevitable. and so I Liam McCabe: Mm. Corey Moen: I just wonder if this this tool from the sketch team is just the next take at that. Again, I think paper and some of these other tools are trying to do the exact same thing where like you the designers can have their comfy little canvas workspace and manipulate and maintain things, but like without them maybe even realizing or they don't really need to care. What they're manipulating is the actual source. It is the actual code. We're just giving you a visual way to do it. And you don't have to leap across plugins and all these things to do it, right? Like that, that is, I wonder where it's going. And so sorry, long-tailed question of like, have y'all been talking about this? Like how where's Figma sits? And like, is your workflow evolving around design systems and some of that? Liam McCabe: So I think the majority of our work in clients still lies more on the creative side of things. So we'll get in a brief to essentially just redesign or rebuild or come up with a website Corey Moen: Two ho. Liam McCabe: design. So that's great because we still just jump into Figma and we muckle that up. Corey Moen: Hundred percent. Liam McCabe: I'm still very code minded. So I'll typically build. our designer Figma site with all the components and auto layouts. I'm not a total creative designer and offset everything. Don't do that. I can't do that. Corey Moen: Yeah. Liam McCabe: I need structure. need order. Corey Moen: Yes. Liam McCabe: Which really should be what Figma is. It could be, again, it's like which route you want to go down, the design system route or the creative throw everything on a page route. But more recently in our workflow, we will do a bunch of mock-ups and they'll probably get to like 50, 60%. And then we'll just start messing around in code and start messing around with interactions just simply because you can generate so many variations, Corey Moen: So fast. Liam McCabe: interactive variations. mean, Figma has added so much now. I don't play around with a lot of what they've added, but I think you can do similar things in Figma. yeah, I've always been a Claude code person for years now. Corey Moen: Nice. Liam McCabe: So I'll just copy everything over and then mock up some interactions and basically pop them up to Versailles or whatever and shoot them over to the client and be like, this is what I was thinking for this component. This is what I was thinking for that component. Corey Moen: Mm. Liam McCabe: And they love that because usually I'll stick a bunch of configurations, switches and tools, parameter fields, Corey Moen: Tweaks. Yeah, yeah, yeah. Matthew: Mm. Liam McCabe: and they can play around. it's like, because a lot of the times, Sometimes the client will have an idea of what they want. But you have to explain to them, it won't look exactly like that because that's hard to actually generate and do properly in code. But we can give you the tooling that will let you play around with it until you like it. And Matthew: Mm-hmm. Liam McCabe: that's so much better now rather than the back and forth like a few years ago. It's like, okay, Matthew: Yeah, yeah. Corey Moen: Yes. Liam McCabe: we tweaked it to what you said. It's like, yeah, but it's still not quite right to the design. Yeah, but it still looks good. Matthew: Because there's like a language barrier between us Corey Moen: Yes. Matthew: and clients. Like we're talking Liam McCabe: Yeah. Matthew: in pixels and hex codes and stuff, and they're Corey Moen: And it's it's this Matthew: like, I don't know, more this direction, a little more that way. And Liam McCabe: Yeah, but now we've given them the nudges to do themselves. All right, so it's it's a dream. Matthew: Yeah. They can do the nudges. And then do you have like a way where like they can just like send you whatever tweaks how they've set them, like the code or just like just the values or Liam McCabe: Yeah, for every panel that we do, I should probably come up with a consistent, like, toggle thing. Whatever you call that thing. You know in 3JS or whatever, you've got the top right panel with all the toggles. Whatever Corey Moen: Yep, yep, yep. Uh-huh. Matthew: Mm-hmm. Liam McCabe: it's called. Corey Moen: Or I always think of like Jay, Jay Hay, the like, you know, famous CSS just wizard and like all of his code pens have the little yeah, the little like tweaks Liam McCabe: Yeah. Corey Moen: panel and toggles and checks and sliders. Yeah, yeah, yeah, yeah. Liam McCabe: But basically each one of those will just have an export button, right? And it's like, okay, when you're Corey Moen: Like Liam McCabe: done, just export that code and we'll take that configuration and we'll use it for this page, that page, this page, that page. Which is great because it means we can build more interactive websites, more fun websites Corey Moen: Mm-hmm. Liam McCabe: without that additional step or two or three steps of explaining what it's gonna look like. Can we have the budget to do this? it will look good. So that has had a huge effect in our workflows. Matthew: Do you think are the clients I mean, how are they reacting to that to the experience of working this way with you? Cause I would imagine it removing some of that back and forth like also like gets rid of f frustration that might emerge just because of the constant back and forth communication. Liam McCabe: They love it, yeah. They love it. And like the fact that you can generate these super, super quick as well, it's like, yeah, it's wow. It's the cliche Matthew: Bien human. Liam McCabe: shock. It's like, wow, this is awesome. Yeah, I'll have a play around with this and I'll get back to you. And we tend to have to wait more on them than us, because it's just the case maybe. Matthew: Always, yeah. Yeah. Liam McCabe: But yeah, I guess that's the kind of... creative design, build type side of things. then internally, we've also just built loads of loads of products internally. Like I imagine Matthew: Mm. Liam McCabe: almost every agency has like, why are we paying for all these services? Let's just build them. Corey Moen: Yeah. Matthew: Yeah. I wonder if ever all the agencies are building like very similar internal tools and it's like it's like this normal set of five, Liam McCabe: Yep. Corey Moen: I mean, Matthew: ten tools. Corey Moen: we talked about earlier commenting, you know, like or feedback tools. I don't know if y'all have gone down that hole, but like I've seen multiple teams and people Matthew: I built my own comment tool for an internal app the other week 'cause I was like I made a request for it from from where I was hosting Corey Moen: Totally. Matthew: this code and then I was like, you know what, I'll just build it in myself rather than and it was it was easier. Corey Moen: Heck yeah, dude. Heck yeah. And like then you can do what you want with it and and have it work how you work or implement it into Matthew: Yeah. Corey Moen: your own, you know, stack and preferences. Like, I don't know. This is such a I I think very exciting time for these things. Yes, it's like Leery and people are like, CS is dead or whatever. Like, nah. Like I it just it needs to evolve and just like, you know, I don't know. yeah. Liam McCabe: Yeah, think when this AI stuff started coming about, I did kind of want to try and automate the whole journey of Corey Moen: Mm. Liam McCabe: just the agency business model because so much of it, so much of it, Corey Moen: Yeah, yeah. Run agency, one button. Liam McCabe: I mean, there are services trying to do this, like agency as a service, just boot up these agents and leave them to it and approve everything. like, it's not going to quite work. anyway, that didn't quite, I mean, it was too much stuff going on. It's like, like a few months or years or a year later, we just started building more individual specific tools for like our content marketing, our lead generation, et cetera. And those have been, yeah, much more worthwhile. I think we're actually using them. But then more recently, very recently, I was working on an SEO agent and yeah, I don't know what happened in the last month or two with these models, but. the spawning of sub-agents and the amount that they can do when you one-shot. So just one-shotting every three to six months, you see the output. It's like, Corey Moen: Mm. Matthew: Mm-hmm. Liam McCabe: the progression is insane. And yeah, very recently I did just another one of those just to see where everything is at. It's like, let's see what it can do. And the latest platform thing that I just one-shotted is just outrageous. Corey Moen: Mm-hmm. Liam McCabe: It's it's overwhelming in... it's feature set and it's UI and UX, but it's not that, it's not bad, you know? It's like, Matthew: Yeah. Liam McCabe: and it's almost gone from the everything app that I wanted to do in the beginning to individual apps back to the everything app. And I'm like, well, why don't we just merge this app and this app into this app now? Cause it's kind of all working and connected. Matthew: Mm-hmm. Liam McCabe: it's like, it's, yeah, it's crazy. it's, cause we, I was doing an SEO agent, AI visibility type stuff, like can you analyze our website, do recommendations, Matthew: Mm-hmm. Liam McCabe: opportunities, and then, well, if we can just hook it up to Webflow, it can just almost do that automatically, just approve it. And then it also had a content part to it. It's like, we recommend writing blogs about this. This will help with your rankings. Like, okay, so we don't need our separate content platform. We'll just add all the features that it's missing over there. So it handles content, SEO, AEO. And then I guess it's not doing leads, thankfully. But I think if you had to have everything in one app, that would just be ridiculous. But yeah, the amount it can do to just manage a website and optimize that website almost automatically is pretty crazy. Like if you wanted to go down that route, yeah, you could just have a self-improving website. But then you ask yourself, well, I ask myself, all these AI optimization, well, AI optimization, it's AI telling you what to recommend so that AI can scrape it. And Matthew: Yeah. Liam McCabe: so it's like... Corey Moen: Meta. Liam McCabe: There is no human involved in that process, which is like... It's Matthew: Yeah. It's writing the content and reading it. Liam McCabe: like, we... Are the agents just optimizing the website Matthew: For themselves. Liam McCabe: for agents? Or is it actual human set? I don't know. And then the agents will be... The Matthew: Mm-hmm. Liam McCabe: agents will be buying stuff anyway if you give it the credit card. Corey Moen: Goodness, yes. Matthew: yeah. no. Liam McCabe: So who are we optimizing for really? Corey Moen: I think Matthew: That is the question. Corey Moen: the one thing I'd add, and then I know we're like, this has definitely been a rad long episode surprise to rapid. It's like the one thing I'd add though is Matthew: Yeah, we need to clean that. Liam McCabe: Yeah, it's alright. Corey Moen: what what you said on like, especially the agency side of like just automate an agency. I think you know, at least back from my own memories and experiences with like freelancing and working with agencies in the past, was like, I think one of the most valuable things, especially in the clients that we ended up working with the longest, was this. What we're doing right now, the like the like let's talk it out and and it's fine if all they don't really care how the output happens, you know, like you know, whether it's from Webflow or whatever it is, like they but I think the ones that are really valuable is when you have those like, you know, the trust building relationship of, you know, trying to solve their the their problems and help them grow their business as a partner more than anything. I don't see that changing. This is who we are as humans, you know, is like This is a big part of how we make decisions on anything is trust and and safety and security and warmth and all of those things. And so it's like never forgetting that side, I think, of all of this new exciting or scary tools is like, let's not just forget that like that matters. You know what I mean? And and and I think genuinely that the people that do ignore that and just go right to the one button solve are gonna be have a rude awakening eventually 'cause it's gonna catch up. Like it's you know, there there's Liam McCabe: Yeah. Corey Moen: only so many clients and people out there that are willing to work with a faceless service, you know, to do something Liam McCabe: Yeah. Corey Moen: as vital as grow their business or represent their brand or whatever. So like, I don't know. Time will tell, but Liam McCabe: Yeah, yeah. mean, mean, yeah, we've never been like fully async. Like there are some agencies that are just async and just use Slack, right? And like, Corey Moen: Mm-hmm, mm-hmm. Liam McCabe: I never understood how those relationships work. Cause I guess that is just maybe a handoff type situation and yeah. Corey Moen: yes, like the the agencies that claim they don't do calls, no meetings, you know, like like Matthew: Mm. It's just a pipeline. Corey Moen: Brett from Design Joy and some of that. same. Hard same. I'm like, I don't know, man. I don't but I don't think I wanna work with the clients that are willing to that are willing to do that, to be honest. I don't know. Yeah. Liam McCabe: Yeah, but I am keen just to try and improve efficiency and productivity and like leave the humans, I hate that we're saying humans these days, but just to do Matthew: Yeah. Liam McCabe: all the fun, the fun, interesting stuff, right? It gives us the time Corey Moen: Yes, yes. Liam McCabe: to mess around with all these prototypes, come up with campaign ideas and actually generate the fun stuff as opposed to... having to do half a day of admin and stuff, a lot of that can just be reduced to like 10 minutes. Corey Moen: Yes, totally. And and feel a little more limitless in that. Like my own experience of this is like I remember saying no to re client requests because of knowing that's gonna take me way too long to build. you Liam McCabe: Yeah. Corey Moen: know, or like I don't even know how to build what you're asking for. So I'm sure I could figure it out. But like again, there's a scary unknown and risk there whereas like that's not gone, but it's definitely it's definitely change, you know, where it's like a lot more is possible and it's like Let's focus on the experience we're delivering and the brand we're delivering. And we don't have to stress how we get there. We know that there's probably one way or another we can find a viable path, you know. Like the irony of even like the last thing I say of that is I mean, no code supply code when I first started putting the directory get directory together, it was like all this stuff I've been saving in Notion forever. And a big part of it was code snippets. I used to have this like huge library of code snippets that I was like always grabbing off the shelf. And it's like, Matthew: Mm-hmm. Liam McCabe: Mm. Corey Moen: the irony of even the last two years of how worthless that is now, 'cause I'm like, why would I why would I ever go back to most of those when I could just have an agent implement it not only probably in a similar way, but contextually in a code base, you know what I mean? Like, yeah, I don't know. It's that stuff evolves and, you know, I think our two Liam McCabe: But that's like, why keep video game disc? Why keep VHS? Because it's fun, right? It's fun to collect this stuff. Yeah. Corey Moen: Yeah, yeah. Nostalgia. Yeah, yeah. Now that you say that, I've gotta go browse through my code symmet library just to rem remember the good old days. Liam McCabe: We've got to stay in the physical world, man. Matthew: You you should print Corey Moen: Fair, fair. Matthew: out those code snippets like and make a book, you know. Like Corey Moen: Yo, have a little I I'm gonna swap Liam McCabe: Yeah Corey Moen: this poster back here to just like a bunch of little code civets, you know. Matthew: Yeah. Liam McCabe: Yeah, exactly. Corey Moen: footer copyright year and Matthew: my goodness. Yeah, all Liam McCabe: Yeah. Matthew: the things. All right, guys. Well, thank y'all for joining. It was a great conversation. Corey Moen: So good. Yeah. Matthew: And I think Liam McCabe: Thank you. Matthew: everyone, let's just go out and let's let's prioritize human connection, right? Whether that's Corey Moen: Yeah. Matthew: agency client or website to visitor, like let's let's let's put humans first and put those touches into everything we do. Corey Moen: All right. Thank you. Matthew: See y'all next time.