From AI‑powered coding epiphanies to unexpected Objective‑C detours, this episode is equal parts tech therapy and hacker joyride. Kalen and David tame verbose Claude 3.7, celebrate Flow’s new multi‑step loops, and brainstorm auto‑styled Shopify widgets that design themselves. They rant about Figma’s “Dev Mode” trademark, debate whether kids should let ChatGPT ace biology homework, and plot a sabbatical‑fuelled walkabout. Grab headphones and join the chaos; the future of dev life has never sounded messier—or more fun.
[00:01:33]
Kalen Jordan: Well, let's hope so. Now I can't, I can't even maximize this, uh, window, but I guess that's fine. Riverside desktop pushed to, coerced me into an update. I'm not sure how I feel about all that, but
[00:01:56]
David: Why did you decide to try using the desktop app for Stability? Hope.
[00:02:00]
Kalen Jordan: yeah, because, I think it was on the web app that one time where it didn't upload my audio. I. And I open it up and it looks nice. Like, it feels good.
[00:02:11]
Kalen Jordan: like the desktop app feels good, like it looks smooth, looks responsive, and it makes sense that like desktops should be more reliable at doing, stuff, data engineering stuff,
[00:02:39]
Kalen Jordan: yeah, and he also said he thought the audio sounded fine. My other buddy shout out to Colby also said that the audio sounded fine, but I'm listening to it and it's like, it's got these, like choppiness is. So, yeah, just
[00:03:02]
Kalen Jordan: That was the desktop app.
[00:03:03]
That was the desktop app. So if it gets screwy on me again, then I'm going to, you know, I, I'm gonna,
[00:03:12]
David: applications. Give that bad boy some more Ram.
[00:03:15]
Kalen Jordan: oh dude, I have a whole, you didn't understand. I have a whole procedure. I close all my windows. Before I have a whole checklist, dude, close all the windows. , Make sure I go to the bathroom. Um, make sure,
[00:03:29]
Kalen Jordan: yeah, make sure my, mic is charged up. Yeah, dude, so the ram should be fine. I mean, I don't want to hear anything about ram from these characters
[00:03:55]
Kalen Jordan: they should be looking into this. They should, somebody should be logging these critical events on their end. I mean
[00:04:02]
David: Don't you have like error detection or like magic? Know what my experience was? Like tools.
[00:04:10]
Kalen Jordan: something, man. Come on guys. 2025. What are we doing here? , the last frontier for me of vibe coating is like a, a Mac app, because, you know what I mean? I've never done a Mac app. It's everything's web-based. It's all web-based baby. But
[00:04:44]
Kalen Jordan: one of the things I've noticed is that like, things that trip you up as a person are, because like I, I've looked at, uh, objective C before and like you look at it and it's just, everything looks so strange. Like, syntax looks so weird.
[00:05:00]
So it's like, that was such a barrier to entry to me. Whereas I'm almost positive that it'll, that won't be a problem for it. Like any weird syntax stuff, it'll be able to figure out, you know what I mean?
[00:06:15]
David: and it'll make a nice little table that like, is super easy to understand and it would be valuable if that information was actually true.
[00:06:24]
Kalen Jordan: Just absolute nonsense. That's so weird. What is, is it a, is it a database format or what the hell is it?
[00:06:35]
David: it's like a, it's a, it's automation for, , building of data, which are really just like a bunch of SQL joins
[00:06:45]
David: that are all stringed together. So the project like helps you string all those together and resolve dependencies and stuff like that.
[00:07:06]
Kalen Jordan: Yeah. We need to get, yeah, dude, we, yeah. We need to get bracelets. That's so funny. That's definitely gonna become a thing. I I'm sur I would be surprised if that wasn't already a thing. Hey, we can add that to, yeah,
[00:07:29]
Kalen Jordan: Um, but, , is it like a code thing? Is it like, would you have a code base for a DBT or is, are you okay?
[00:07:39]
David: have a code base. It's like, um, so every file, like every code file is a dot sql file. So you're just doing SQL stuff and you can reference, it's like a, a hierarchy of. Like you run a SQL query that creates a table somewhere, and then you can run a query on top of that table that was created to get it a little bit closer to something that is easy to grab information from.
[00:08:03]
Like, for example, in Shopify, the base order table that comes in when you do an extraction
[00:08:16]
David: steps in this would be like, okay, grab the orders table and then grab the tags table and put the tags directly into a new table with the orders. And then you've got like an orders table that's got tags on it already so you don't have to do the weird stuff
[00:08:44]
David: Yeah. It's got all that stuff. So you have to figure that out. Every time you change a version, you gotta figure out like, okay, this thing depends on this other thing. I gotta update that. So I was in health for a few days this week,
[00:09:05]
Kalen Jordan: it's like you're running from of the flame. It's just like there's flames in the background and you're just running, running past them.
[00:09:13]
David: you never want to shave, you always got to
[00:09:18]
Kalen Jordan: I always forget what yak shaving means, and then somebody explains it to me and I go, oh, yeah, that makes sense. And then I immediately forget it
[00:09:28]
David: you need to shave the yak, but then you forgot that you need to buy like a new razor for your shaver. So you go to the store and then
[00:09:42]
David: though you just wanted to shave the yak.
[00:09:43]
Kalen Jordan: Right, right, right, right, right, right, right, right. Yeah, that's, yeah, that's, that's, it's, it's like you can be doing something and you're like, so productive and you're humming along and then you go, oh, I'm gonna do this other thing. And you kind of know high level how it's supposed to work, and you're like, great.
[00:10:01]
And then you start doing it, and then all of a sudden you're in, you're, you're four levels up in that stack
[00:10:33]
Kalen Jordan: that's cool. So you can use this DBT stuff to, to wrangle all the data that's in NetSuite, or is this after? It makes it into your Google big table, BigQuery.
[00:10:44]
David: It is after it makes it into BigQuery. So they come
[00:10:55]
David: like, okay, obviously we're gonna want to our orders based on certain tags and so we have to join the tags on. And so it's just like a bunch of convenience stuff.
[00:11:06]
David: yeah. And, it sounds cool, but it takes forever to get it to actually work, but now it's working, so I'm excited.
[00:11:14]
Kalen Jordan: Nice dude. And then you just, like, you're just in your code base and you go, okay, let me, add this little dependency, this little sub table or meta table or whatever. And then it just, it just deploys it to the cloud and everything. And like,
[00:11:32]
David: it all just becomes tables and BigQuery. It's
[00:11:47]
are there any, like, any Nice little things you're able to like, query on that would be completely impossible? Doing the regular APIs that like, and any like nice little OTs that you've been able to slurp outta there?
[00:12:03]
David: OTs would be, there's a, so the prebuilt package, it's really nice 'cause Tran who does the, extraction of the data and puts it into BigQuery as a first step. They're the ones who made these DBT packages and the, the Shopify one they made. Already does a new versus repeat order calculation based on whether or not this is the first order the customer has placed
[00:12:34]
David: is another one though. Like from the API like, it's not hard to get tags for a specific order, but this scenario you're looking at like all orders and so you want to do like all orders that have a certain tag go into this group,
[00:12:59]
David: Yeah. Hopefully I'll be able to make some real progress on this next week now that it's, now that I'm out
[00:13:04]
Kalen Jordan: Now that's outta dependency. How? Yeah. So the, the package, is it a node package or is it some other kind of
[00:13:13]
David: Like there's a whole, like DBT has this package hub kind of thing, and it looks a lot like Node. Like you go to the thing, so like I can even just search five trans slash Shopify and the DBT package hub will come up if
[00:13:33]
David: it's probably all built off of like a GitHub repository. Like notice,
[00:13:38]
Kalen Jordan: Oh, nice. is there a local command line or local environment where you're running stuff or is it just like you link it up to you, link up the package to your Google, big query and it slurps it in from there?
[00:13:58]
David: You run the command line to resolve the dependencies. It does stuff. And this is the part where chat, GBT was like sending me on a wild goose on It was like, okay, so there's two modes to this, uh, CLI, there's a local and there's a cloud.
[00:14:14]
And then I was like, it doesn't seem like mine is either one of those. And then it said, oh, there's also a hybrid approach and this is how
[00:14:43]
David: finally started working. So now when I run, so I can make code edits and vs code and then run this command line and it'll send everything up to the cloud, run it all in a development environment, and then create those intermediary tables and like a schema in BigQuery.
[00:15:18]
Kalen Jordan: Well, basically you can go, Hey, sync customers and customer addresses, and they're just little check boxes and sync products and variants and product media.
[00:15:27]
And you can, and then they'll just sync all that in for you. For a price. And then also you can create your own models and things like that. And, but so previously with this B2B project, I've been working on, I've been trying to avoid using, uh, I've been trying to do everything native Shopify.
[00:15:45]
So instead of, so like, for example, we have the concept of like territories by state and country and zip code, and we put those into meta objects, for the territories and like zip code mappings, we have like 57,000 zip code mappings and meta objects, which is like, it feels cool to be doing things natively, but then you run into like limitations where like you can't filter on something you wanna filter on or you can't join on something or whatever.
[00:16:15]
So we've been, migrating some stuff over. So it puts a bigger dependency on, on Gadget, and it's like you're, you know, you're like a little further away. What made me think about is what you said is like, you're a little further away from like, native,
[00:16:30]
Kalen Jordan: but it, you know, it, it lets you do some things like,
[00:16:34]
one of the things, there's a dashboard where the, the merchant can manage customers and in their previous site they could sort on like a last login, field. But we made that a meta field and it's indexable, so you can search on it, but you can't sort on it. So like we, there's no sort feature.
[00:16:53]
And so moving that data into, into gadget now, you can sort on any field you want, you can search on any field you want. So it kind of feels like cheating though, you know? 'cause like there's something like fun about doing everything native and trying to get the best solution possible.
[00:17:11]
But sometimes it's also a, a, a vanity thing. Like, sometimes it's, it's like really important to keep things super native. And then other times it's really not that important. And like, knowing the difference, you know.
[00:17:36]
David: the difference between using the gadget like model stuff versus like, I know you were a big fan of storing stuff in Airtable. it kind of like that?
[00:17:45]
Kalen Jordan: Uh, yeah. It's just, uh, it's just baked in, I think they do Postgres under the hood. so it's yeah,
[00:17:52]
David: APIs to get at those models, I'm guessing. Okay.
[00:17:55]
Kalen Jordan: yeah, yeah. The APIs and there's like front end components that, you know, you can just flip, flip up a flip up. You can just flip up a, like a grid of a, a, you know, on a certain model and like a form and like bulk actions on the grid and things like that, that, you know, little convenience things. Um, yeah,
[00:18:19]
David: at some point Shopify will buy gadget and it'll
[00:18:32]
Kalen Jordan: And I was like, which honestly, I like, I was happy to hear that. Like, I, I, I think it's so rare these days that people don't want to be acquired, but you know, things almost inevitably go downhill.
[00:18:46]
I mean, even checkout blocks as like, sort of like the, if nothing else, the support is not as good, but it is free. But like, I mean, almost everything sort of goes a little bit downhill, uh, when it gets acquired in certain ways.
[00:19:04]
David: people who they acquired on like different projects and stuff like that. I.
[00:19:08]
Kalen Jordan: Yeah, yeah. Although probably Shopify more so than almost any other company does a good job with acquisitions. I think. I want to say that I can't think of any specific examples.
[00:19:20]
David: who else did they acquire? They acquired
[00:19:27]
Kalen Jordan: I mean, they acquired stuff left and right. I, I saw they acquired a search thing recently.
[00:19:34]
Kalen Jordan: re Right, right. Remix. And I don't know enough about like the internals of remix to know whether that ended up like being a, a big benefit to remix or not. I think remix now is React Router. It's like React. I keep seeing people post about this.
[00:19:54]
That remix is just a React Router package or something like that. I don't really get it, man. I just, I follow the docs. I try to get the shit to show up on the screen. And that's about, you know, that's about the extent of it. But, um,
[00:20:08]
David: that guy who was working on it is still doing remix stuff. I think his
[00:20:20]
Kalen Jordan: yeah, but it's funny that you were, doing the stuff in, in chat GBT, because actually I think you were one of the first people that, when I mentioned I was using Cursor with like, uh, chat GPT, you were like, oh, well have you tried Claude? And I was like, how do you use Claude?
[00:20:40]
What is Claude? How do I even, I don't know how to use it. And then I started hearing, you know, repeatedly people say, oh yeah, Claude is generally better for code. And now like, that's all I use in Cursor. Like, the idea of using chat GPT in, in there is like, yeah, I have zero, I have no confidence in it.
[00:21:14]
David: I'm, I'll make my way back to Claude. Don't worry.
[00:21:17]
Kalen Jordan: Now the images stuff is sweet. Dude. I've been having a good time with that. Um, yeah. I've been using it for like, little icons and, and I did some like, yeah,
[00:21:31]
David: generation, like right before that came out.
[00:21:34]
Kalen Jordan: yeah, I had some trouble doing sprites actually, because it, okay. it did a nice little Sprite of the thing I wanted, like I wanted to, make this video game of Joe Rogan punching people. And I, I uploaded an image and it made a nice little Sprite of it. But the layout of the Sprite, yeah, I guess Sprite has to be perfectly, you know, each frame has to be the same exact width and height, and then the character has to be in the same position within the frame.
[00:22:07]
And it couldn't get that right. Like the, the, the art, the artwork, the pixel artwork looked great, but it was like aligned, funny. , But it was close. It was like, oh, I could finally get this thing going. 'cause, yeah, but like I've done some like little icons and like little website illustrations to, to pop into, an app or a web app or something like that.
[00:22:38]
David: uh, I was making some slides the other day and I was like, I'm doing this slide about this thing. It was, , where our team is using AI to help us code because the, the Toby, mandate to use AI
[00:22:59]
David: posted some of that memo in our staff meeting yesterday. Like, we gotta do AI stuff. So I was talking about what we use AI for It generated an image for my slide, and it was like a perfect little laptop in the middle, and then some like flow charts about how it was helping with scripting and every word was perfect, which was
[00:23:21]
Kalen Jordan: yeah. That's so cool. Yeah, it's like there's some like fun, like trendy use cases, but, which, you know, are cool, like the Ghibli stuff, but,
[00:23:38]
Kalen Jordan: a sub button icon that like you'll actually use.
[00:23:42]
Um, so like that's pretty sweet. I, I'm, I'm really,
[00:23:46]
David: to ask you if you, there was a another image thing that came out, I think from Google the same day as Open AI released their new image thing,
[00:23:58]
David: Gemini studio or something. And I saw people actually creating, sprites using that, but I don't remember what it was.
[00:24:06]
Kalen Jordan: oh, nice dude. I keep hearing stuff about Gemini. I think I've seen a couple people say they switched from Claude to Gemini 2.5 for code. I've heard like those types of things and I still just, I just kind of glaze over it. I'm kind of just like, I, I already have my thing that's working. I don't have time to like, even though it's like sort of trivial to just try it out, it's weird how you just kind of go like, ah, I don't wanna deal with this right now.
[00:24:39]
David: thing with the different OpenAI models. If anyone ever asks what model I use when I go to chat, TI, I'm just like, I don't know. I don't ever use that switcher. I
[00:24:58]
Kalen Jordan: Yeah. I was really confused the other day 'cause I was like, okay, I was trying to figure out whether GPT-4 oh was better versus oh 3, 0 3 versus oh four mini. Like, I don't know. Oh four oh mini versus oh three mini versus, oh, I don't know, man, it, it was wildly confusing. Um,
[00:25:29]
David: one is better than four. Five. It is very confusing. And they know it.
[00:25:34]
Kalen Jordan: Yeah it's a absurdly confusing, there must be reasons for all this is, I think there's like significant differences in the, under the hood and how these models work. Whereas like, we, we just think of everything as like a linear thing. Like, okay, it's getting better, so you bump the version number, but they're like, no, no, we're like, it's like a different things, you know, somehow, I dunno.
[00:25:59]
David: Yeah, I don't care. Just like let the scientists deal with that. Just give me the best one, please.
[00:26:04]
Kalen Jordan: Yeah, yeah, yeah. But I finally started using Claude three seven again recently. 'cause I had, in cursor I had switched to three five because three seven would do way too much all the time. Like, you'd be like, fix this button. And it would be like, okay, boom. And then I'm also gonna go over here and, and, and like optimize your database.
[00:26:27]
And it would just, it was absurd. And then I saw somebody, so then I was like, okay, that's fine. I'm covered with 3.5. It's kind of dumb, but I'll go through and I'll fix the things it does wrong. And it's still is like helping me, but I know how to use it. And then I saw somebody tweet it. They were like, I, they said they use this prompt that says like, check with me before doing anything and don't do anything other than exactly what I'm telling you.
[00:26:54]
And so then I started using that prompt. And all of a sudden it was like perfectly not, I mean not perfect, but like significantly better at just doing the thing I was telling it to do. , And it's like so much better than three five, and it was, and it's faster than three five, which is weird. And then in the last few days I've stopped putting that little preamble in I think they fixed it.
[00:27:23]
Kalen Jordan: they're, whatever system prompt you're using, they're getting it to avoid doing too much madness. But it kind of does, it's still kind of proactive, like it'll do some stuff, but most of the time I'm like, yeah, okay, cool. I wanted that little extra thing even though I didn't say it.
[00:27:38]
And dude, it's just been freaking great. Um.
[00:27:42]
David: it again. 'cause that was the problem that turned me off of it a while. Like when I first tried it, was it just went way too crazy
[00:27:49]
Kalen Jordan: Okay. Yeah, dude. Yeah. And I've used it on some theme stuff lately too. , Like some, some fixing, some theme issues and, and spelunking through some CSS and some JavaScript. And it's been, it's been pretty, pretty decent.
[00:28:07]
David: Yeah, that's what I would try it on. Maybe I'll give it a shot this weekend.
[00:28:10]
Kalen Jordan: nice. Yeah, actually this one project that where this, these weird custom bundles with all this over complicated JavaScript stuff. I dunno if you remember that. I had mentioned it, uh, before, but I would spend like, , they'd have like a relatively simple bug. Like, you know, when we tried to add this item to the m it throws an error and it would take me three hours of digging through this insane JavaScript.
[00:28:39]
And so I was praying that they would, like, I even told 'em, I was like, Hey, this really might not be a good fit for me. They're like, no, no, no, we really wanna, and I was kind of just like hoping they'd stop asking me to, to work on it. It was 'cause of how painful it was. And then just recently, they got back to me and I, and I worked on this one thing and it was like, I.
[00:29:02]
It was almost fun this time , because it like figured things out and um, and
[00:29:09]
David: an agency or something that wrote all this JavaScript.
[00:29:12]
Kalen Jordan: I inherited from this other agency. And I will say that like, while the code to me is super over complicated, like there is a, like a logical structure to it, it's just overly complicated. So like, as a person when you're trying to understand it, it's like really slow.
[00:29:32]
But I feel like the model was able to understand it pretty easily. ' 'cause You know what I mean? Like, there's a structure to it. It's almost like when you have like, something inherited from something else, inherited from something else. And it's like you have some factory pattern or like, and, and it takes forever to understand it, but like there is a method to the madness.
[00:29:55]
You know what I mean? So I think it was just.
[00:29:58]
Kalen Jordan: Yeah. Yeah. And like every time I came back to it, it didn't like, it never fit. But, yeah, I, I almost was like, man, I wouldn't mind doing more of this. Like, it was kind of fun and because it went so fast, you know, I was like, I was able to build, you know, some good time on it.
[00:30:36]
Kalen Jordan: oh dude. I think I have a max of like. At first, I was on the free tier, and then it was like, it starts limiting you to slow queries. So then I bumped it to like $20 max, and then I bumped it to 40 60. I think I'm at $80 per month max right now.
[00:30:53]
So I haven't bumped over that yet, but like the Claude terminal is like 10 times more expensive, and I haven't been using that a whole lot, but it's, it's noticeably,
[00:31:06]
David: stuff? It's just like there to hold your hand while you're in the terminal doing things.
[00:31:10]
Kalen Jordan: well I think it like spends more time, understanding the whole code base and I think it probably has like more context it's using, and it's a pretty cool interface. Like the terminal, like, it basically will show, it'll, you tell it to do something and it'll show you a diff and it'll say, do you want me to make this change?
[00:31:33]
David: a, you're doing stuff in the terminal. It's like still code, project based.
[00:31:37]
Kalen Jordan: yeah, yeah, it's doing the code, which is kind of weird for me. Like, I like being able to look at the code in an editor and you know, scroll around and things like that. So it's a different, like, way of doing things. But last, I used it, I did think it was like a lot better at like, doing things across the code base in a really smart way versus like, Hey, look at this file and this file and do something in Cursor.
[00:32:17]
Kalen Jordan: yeah, it's a, it's a weird concept. And then I think chat, GPT just launched a terminal thing called Codex, which I haven't even looked at yet. I think they just launched that the other day, so I gotta take a look at that.
[00:32:32]
David: Dang, I gotta catch up on some stuff.
[00:32:35]
Kalen Jordan: dive in, dude. Jump into it.
[00:32:39]
Kalen Jordan: Yeah, I've been doing a bunch of Polaris and, , in the past I had been really slow at, you know, Polaris, do you know, building Allo app interfaces and even using gadget for like, sending data back and forth. I was extremely slow at it, but dude, it's been j I've been jamming on, on that stuff and it like now I did have a foundation in place of existing Polaris stuff.
[00:33:06]
With grids and buttons and modals and kind of every little piece, like every component you would need. So I could point it to like, do the, create this similar to this page. And then it would just, I mean, it just, boom, create a modal, you know, make a, just make a button, do this and do that, and do this and that, and it, and it's just nailing it, dude.
[00:33:36]
Kalen Jordan: No, it just figures it out like, like zero, you know? And even like weird hydration, react things, which would always throw me for a loop, paste the error in it. It pretty much like figures it out. It's cool 'cause it's like now I can just work in react. And Polaris and stuff without, really having to be stuck in the weird shit and just, you know, thinking, you know, working on, on a higher level of like, okay, I want a thing here and a thing there, and I want this to do this and this data to talk to that.
[00:34:10]
David: I saw Shopify has an MCP for like GraphQL queries. I wonder if they have something similar for helping models understand Polaris or if they just do.
[00:34:23]
Kalen Jordan: Yeah. I am not sure. I think they, they sort of just do, because like you, you know, you like, you just say, Hey, you're importing a text component in a block stack and a button and a thing. And so, and sometimes you have to point it to like the Polaris stocks to figure stuff out. But most of it is, it kind of falls into a standard like react component framework.
[00:34:46]
So it, it can kind of just put those building blocks together. But, dude, the MCP thing, I'm so lost with it because the last time I used it, it like, it won't automatically call the tool. Like it, like I was asking it like, Hey, I need to do this graph, this GraphQL problem with this query. And I would've thought, you know, it would go, oh, let me go check the GraphQL Shopify docs.
[00:35:13]
It's like the only MCPI have installed. It didn't do it. I had to specifically ask it, like use, you know, to use the Shopify MCP for it and then it would do it, but then it like asks you to confirm every single time it does a, a tool call where I'm like, just do it. You know, just do it automatically.
[00:35:29]
So I'm fine with me. , But I don't know. And I think there's a way to have them go automatically. I've seen a couple people say that, but I don't know how to enable it. And then even then. It still didn't correctly fix the particular GraphQL issue that I had. So I'm super lost with it right now.
[00:35:49]
David: Yeah, I haven't tried it yet, but it, it seemed like it would, because like the, the whole thing with MCP is it has the entire GraphQL schema, so it shouldn't go outside of that.
[00:35:59]
Kalen Jordan: Right. Yeah. I, yeah, like I, when I saw them release it, I was like, this is amazing. And I still think maybe I just don't have it configured right. But, um. Yeah, but like this, so this, uh, dude dev dev with Alex on Twitter, he launched this thing called Dezi, dot dev, DEZI. it's a UI builder to build Polaris layouts, and it's really nicely done.
[00:36:26]
It looks similar to like the, the theme editor where like in your left column you have, you know, blocks and then you click on 'em and you can just do stuff and then you can export it to react code and drop it in,
[00:36:41]
Kalen Jordan: is like, is super cool. I just, like, I haven't logged in and checked it out yet, but like, my thing is, I feel like you're gonna be wanting to do all of this stuff in, in your AI editor versus doing it manually in a tool.
[00:37:00]
Kalen Jordan: I just think that like the, the default is shifting for any kind of a tool to where you're gonna wanna do it in, in your code editor.
[00:37:35]
Kalen Jordan: Yeah. Yeah. It's very, very well, well done. And he also has, I think, a project called Polaris Components or something where he's got some components that he's created for like little types of widgets that you would want to have in your app that, that aren't like out of the box widgets.
[00:37:53]
So,, that stuff is pretty sweet. yeah, dude. , Flo multiple steps in a loop. Did you know that, that Flo added support for multiple steps in a loop?
[00:38:03]
David: I did not. That is very exciting.
[00:38:09]
Kalen Jordan: Yep. That's, that's kind of a big one. I remember thinking when I first looked at flow, like that's kind of a gnarly limitation
[00:38:18]
David: I don't remember how we worked around that. Maybe we like, uh, I don't remember.
[00:38:23]
Kalen Jordan: Yeah, it's, tricky. Sometimes you can loop over the thing, do a thing, and then loop over the thing and do a thing again like in, in serial kind of some, sometimes you can do that.
[00:39:07]
David: but you're looking for like a, a way to auto detect that or something?
[00:39:11]
Kalen Jordan: Well, yeah, I was just thinking that like, you could maybe just generate, which by the way, when you do that type of stuff with apps, do they tend to integrate well into the theme?
[00:39:26]
Kalen Jordan: Yeah. Yeah. I haven't used a ton.
[00:39:29]
David: curious like, what is the, so like, I, I guess the, in the, the case where this is important is you wanna make some change to the front end, like, and you, and you have to know what the structure of the swatches looks like or something like that.
[00:39:48]
Kalen Jordan: Yeah. Like, because what I'm imagining is like let's say you have a, wishlist button or, so, you know, there's an app that it adds a wishlist button and it's like, normally you gotta go in into this style settings and then you gotta go, okay, here's the background color for the things and here's the, and there's a whole bunch of colors you can customize and fonts and stuff.
[00:40:12]
But I feel like that type of thing, like I could just imagine if I was doing it individually in cursor and I had the theme code base that it would probably one shot it, like if I said add a button to this page following the style of this other page, that's the type of thing it would do.
[00:40:31]
Like, I think really easily. Then boom. Like there, like now, instead of this whole rigmarole of like, and for apps, it's gotta be super hard to like you're saying, if they can let you select from a bunch of themes and now they gotta go out and support 200 themes and figure out how to do all that.
[00:40:54]
Versus I can imagine, a world where like you just, , inspect the, now I don't know how you'd inspect the theme. I don't know if there's API access to the theme Exactly. But you just inspect that sucker. Give it a decent, low prompt. I Think if you could give it a prompt, like where you said, Hey, add this thing here to this liquid page, make it similar to this other thing on this other liquid page, I think it would nail it.
[00:41:21]
David: Probably like the, the thing that's important is making sure that you have the right CSS selector for
[00:41:29]
David: the front end. I mean, like, I, I don't know if this would be interesting, but what if your wizard was just like, okay, go to the wishlist page and , click on the wishlist button, and then like, now I know what the wishlist button is or something like that.
[00:41:48]
David: But yeah, if you have access to the theme, then you have access to the JSON for the templates for each of those pages. So now it would still be nice to be able to like grab the, actual resulting HTML and try and figure it out from there.
[00:42:06]
Kalen Jordan: Yeah. And maybe it could be a thing where like , you could, Put your own prompt in as the user and go, Hey, do this and that, and this and that, and then run it and then see how it does. And then maybe go, oh wait, let me, let me prompt it a little differently. And then, and then boom, you get it to just style that widget, like just the way you want it to.
[00:42:29]
David: comfortable your customer is with doing that.
[00:42:32]
Kalen Jordan: yeah, yeah. Or I could see that being a part of like an onboarding flow where like as the app, I have somebody onboarding and I spend a couple minutes, you know, prompting it to get the style right. Like that would just be cool because so much of the time when I see apps, like they just don't ever, in, especially if you have a custom theme, or a highly customized theme, they just don't integrate.
[00:42:55]
I mean, it's just impossible for them to like integrate really perfectly into the theme.
[00:43:15]
David: something that AI can negotiate.
[00:43:18]
Kalen Jordan: right, right. Yeah, that would be cool. And then the focus could be more on them building out like the APIs for the front end to talk to the app, and then the actual, like front end markup could be like a little more flexible. ,
[00:43:35]
Kalen Jordan: I could imagine doing it, like, I want to kind of try it now, like for an individual app in an like an individual theme that I have access to. I think it would be pretty straightforward, but then , like productizing it and putting it into like an, like a, a widget in the admin and figuring out how to access the theme, all that kinda stuff.
[00:43:54]
That would be tricky. That'd be kind of gnarly.
[00:43:58]
David: though. You keep finding these like ideas that you could license to other app developers.
[00:44:07]
Kalen Jordan: I guess so. We gotta figure that out. Damn dude.
[00:44:13]
Kalen Jordan: Gotta figure it out. It's tough, dude. I keep coming up with little things and working on little things and then, you know, you get a little traction, but then it's like, not enough. It's, dude, it's tough, man. People are picky. Like people are picky when it comes to Shopify apps, dude.
[00:44:30]
Because the bar is really high and there's so many options, you know, so,
[00:44:36]
David: And like everything, like when you're working with shop, like I feel like people who are using Shopify are just spoiled.
[00:44:46]
David: working through this DBT stuff and like trying to get the CLI to work would never happen. Like I, there wouldn't be a case where I would just be missing some weird configuration because the CLI for Shopify would be like, Hey, stupid, you're
[00:45:21]
Kalen Jordan: copywriting, the phrase dev mode. I just feel like it, it's like the whole ethos of, developers building the internet. You know, like the internet emerged from a bunch of like altruistic. I mean, I know there was fucking ARPA net and military and stuff like that, but like open source, developers just building stuff and it's like little by little, all these things built on top of different layers and open protocols and all this kind of stuff, and like open source packages for this and that, and then now the phrase dev mode is copyrighted.
[00:45:59]
That's so stupid. It's just so against like the ethos of what built the internet is kind of how I feel.
[00:46:10]
David: I agree. I bet you this is what happened. Figma almost got acquired by Adobe last year, right? Or a couple years ago.
[00:46:19]
Kalen Jordan: Right? Oh, it got, it ended up, yeah.
[00:46:44]
David: But I also heard that Figma, I mean, I love Figma. I use Figma every day because my creative team just like so fast at putting stuff in Figma, and we just
[00:47:01]
Figma is a cool, it's like a cool tool, you know? I mean, it's like there's things on the internet that are like old and like crufty and like,
[00:47:13]
Kalen Jordan: there you can just see the sleazy business guy behind it all. And then there's like things that are cool that are like in line with like, kind of the heart of like the cool internet.
[00:47:27]
Kalen Jordan: stuff and the not cool stuff. And like Figma is one of the cool things, but it's like, it just, you know, it's like you said, man, they got too many fucking lawyers involved. Um,
[00:47:37]
David: I don't even know what dev mode is in Figma though. What is it?
[00:47:41]
Kalen Jordan: I don't know, dude, I think it's like, you click, I mean, you use Figma, so I, I'm guessing it's like you click a button and then you edit code directly or something like that.
[00:47:54]
David: open Figma tab and see if I can click it. I
[00:47:56]
Kalen Jordan: God. And then just the tone of the email, it was like, Hey, we saw you, uh, da dah, dah, dah. And, and it was like, yeah, we, you know, while we,
[00:48:23]
Kalen Jordan: yeah. oh. Uh, so shout out to, Taylor page again so he has a site called Shop Info App. And, it lets you look up a domain and it tells you what theme they're using.
[00:48:37]
And it also has like a theme directory where you can see a list of all the themes, Shopify themes. And he recently, , posted that he added like a price tracking feature where it, it checks, it, tracks like price changes, price increase or price decrease. And if you go in the theme directory and you look at the prices, it'll, it'll be like this price increased by whatever percent in the last 30 days.
[00:49:05]
Kalen Jordan: yeah. So I thought that was, a cool little feature.
[00:49:09]
David: An Amazon prize tracker situation.
[00:49:12]
Kalen Jordan: Yeah. Yeah. The only thing that would be cooler is if it tracked, like, which themes were like trending, you know, based on like how many sites we're using them. You see like a, a chart of like theme trends and then you also have to kinda like filter that. You need some curation on top of that. So it's not just showing like, the absolute numbers.
[00:49:38]
'cause like, I'm almost willing to bet that whatever top themes have the most number of users are somehow probably not good. There's probably like, there, you know, there's probably like a, like ones that like better sites use, you know, uh, I don't know, but like, if, if.
[00:49:57]
David: kind of bad for my domain because it's telling me that my theme version is like four versions old. fine. we have, we've broken away. Okay. We've, branched off. worry about it.
[00:50:13]
Kalen Jordan: I think this is one of the very first things we talked about was like I said something about like a theme upgrade. , I was like, Hey, , I think this thing showed me that you can just upgrade your theme. And I was like, Hey. I was like, are you gonna do that? You're like, I, I don't even know what that would mean.
[00:50:31]
At the time, I didn't really understand it, but now, now that I've, worked in like, where your customized themes, I'm like, yeah. Like it's, forget about it.
[00:50:41]
David: I'm not going to try and do that. I'll take some of the cool. from the new one, but I've
[00:50:55]
David: There's no, like, security issues. It's fine.
[00:50:58]
Kalen Jordan: right? , And like. This is one of the things with the, with Shopify that like could, feels, like, should be, there should be more attention paid to like how do you, like a Magento at least had the mechanism of like theme inheritance and like local overrides. , Like there should be something like that when you're working on a theme.
[00:51:16]
There should be something like that to some core mechanism for like a local modification to it, to a theme that, so that the theme is still upgrade safe. And even when Magento had those, it still was, it's still, it reality wouldn't quite, nothing was ever quite totally upgrade safe. Like there's always weird stuff would come up.
[00:52:01]
David: exposing myself, I don't upgrade my theme. Sorry.
[00:52:06]
Kalen Jordan: That's the thing I suspect anyone that's like, you know, got a real business that is doing real custom stuff, I, I don't think anybody does.
[00:52:18]
David: A merchant to update their Magento version was always like. You know, we have to do this because of security. It wasn't like, yeah, you're gonna get new
[00:52:38]
David: okay, I'm gonna spend a week upgrading this theme and what are we gonna get out of it?
[00:52:42]
Kalen Jordan: right, right. How did you like that little luxurious, uh, chat GPT upgrade for your site?
[00:52:50]
David: The new font was the biggest thing I noticed, but I really like how it was like everything was centered was very different
[00:52:57]
Kalen Jordan: Yeah, I kinda liked it, dude. I was kind of into it.
[00:53:01]
David: Oh, I forgot to show that to my creative director. I'm gonna have to, and see what, what he
[00:53:05]
Kalen Jordan: I need to tell it to do a better desktop version 'cause it was kind of a random, uh, mobile version. But like I fed in the desktop site at, to the prompt and then it returned back a mobile version.
[00:53:29]
David: the mobile. 'cause we get a lot more customers there.
[00:53:32]
Kalen Jordan: Yeah. I do think if it, like one of the big things it has is like the product shot didn't have a white background. It had like a transparent background with a nice little shadow.
[00:53:43]
Kalen Jordan: think. I feel like having that looks pretty slick. Like,
[00:53:47]
David: Like the, the product itself has a bit of a shadow, but it's just the same background.
[00:53:52]
Kalen Jordan: Yeah. Yeah. And then it can just sit on top of the back of background of the site. And then that probably wouldn't be too hard to automate. Like you could somehow probably remove the backgrounds of the images on in bulk.
[00:54:07]
David: Maybe. I, I gotta figure out, , we've been talking about like potentially cropping all of our images and that's just a ton of manual work for all of the retouchers.
[00:54:28]
Kalen Jordan: I think the video AI models are getting scary. I saw, I saw some, uh, some videos recently. I think one's called King something. Man, they're getting so good.
[00:54:41]
David: Yeah. I don't know how people are making videos out of images now, but there was, , so we, ran the talent show last week at my kids' elementary school.
[00:54:58]
David: she voiceover all of the, you can tell she created images and then added movement to them and then sequenced them and then like, put in a voiceover.
[00:55:08]
David: the kids are doing the AI stuff, dude.
[00:55:11]
Kalen Jordan: That's so cool. My, my wife is really against the AI stuff, so like, I want my kids to be like going all in on ai, like for schoolwork and stuff like that. But
[00:55:27]
Kalen Jordan: my wife is very much against it. It's so, uh,
[00:55:32]
David: like environmental impact or like stealing
[00:55:36]
Kalen Jordan: It's just that she's very big in like a classical education and like handwriting and reading the classical literature. Like, uh, so anything tech forward is sort of like, it goes kind of against what she believes in. And then AI being like the most extreme example, you know, so like the idea, I thi.
[00:56:01]
Kalen Jordan: Yeah. Yeah. Like, whereas me, I'm like, I think they should be using it. You know, like, I don't think spelling is that important of a skill. Like I think,, you should probably find other things to, spend time improving the skills, you know, like higher value skills.
[00:56:17]
David: that's a place where it's like it's obvious to everyone that it's going to have a major impact, but it hasn't really been well integrated yet because like. the way I use AI most is someone who can teach me things or like
[00:56:53]
Kalen Jordan: Yeah, like in the, and like my daughter was, studying biology and, and sh and so, you know, she has the textbook and she's like, reading through the textbook and I was like, dude, like what I would do is just have chat GPT, like tutor you through it and like maybe, maybe start with a question, have it explain the question and then maybe just have a conversation with it back and forth.
[00:57:16]
You know what I mean? Like, I, that I would completely go that direction. But she was like, no, I'm gonna do the book and, you know.
[00:57:25]
David: If the book had like a PDF version, maybe you could load it into the context or something
[00:57:33]
Kalen Jordan: and things like that. Like, you know, high school biology, genetics, those types of things. Like it knows those things already. Like those are very well dialed into its training data. You know, it's not like teach me about Shopify, Polaris is, you know, latest version. You know, it's like