[00:00:00] David: By the way, I just realized that your notes are written on the back of what looks like a multiplication table homework

[00:00:07] Kalen: assignment. That is correct. That is correct.

[00:00:13] David: Oh man. I have flashbacks from like fourth grade having to go through those as fast as possible's.

[00:00:19] Kalen: So

[00:00:19] David: funny. You're

[00:00:20] Kalen: just staring at it. You probably were able to tell by zooming in on it,

[00:00:24] David: right.

[00:00:26] I, I was just looking at it. I was like, why are all these boxes on here? Must be some interesting, like grid paper that Caitlyn's got. Nope. Just stole his daughter's homework and wrote notes on the back.

[00:00:37] Kalen: Just lifted her homework. Oh, she

[00:00:41] David: doesn't need this.

[00:00:42] Kalen: That's so funny.

[00:01:31] With my, middle daughter, Juliet now, I used to just make her laugh like all the time, and she's actually like really funny. Now. I'll try to make her laugh like that, like a dumb bora, you know what I mean? Like a bora voice. And it used like I used to reliably be able to get her laugh with any funny voice, like any time, right?

[00:01:51] And now, dude, it's so hard to get her to laugh now. Because like she only laughs it like something's like actually funny, you know, type of thing. Oh, that's so sad. And, but I'll like, I'll start it and I'll commit and I'll just keep going. Yeah. Like I'll be at the dinner table and I'll just be like the Borat, I'm just kicking it up to 11 for like a solid like eight minutes.

[00:02:13] Oh yeah. And she's just staring at me going

[00:02:15] David: for way too long. Is itself funny?

[00:02:17] Kalen: Yeah, yeah, yeah. Exactly. I was like, that's my only chance now is I have to do the going for way too long angle. But even that, I've like dried that. I've dried that well up, so. Oh, did you get

[00:02:33] David: The message I sent you through your other daughter?

[00:02:36] Kalen: I did, yes. Yeah, she told me, she came in and told me right away. She's like, he says, what's up? Just like this. It's funny because I was listening to the podcast. We were talking about the robotics thing and you were talking about your son. And you normally just have this like blank generic spot in my brain when you talk about like your son.

[00:02:57] I was hearing him in the background the other day when I walked into the living room when Livy was doing the lesson. And then when I was listening to the podcast, I was like, oh, that's the dude he's talking about. He said, mad man. He does not wanna play piano, but we're trying

[00:03:12] David: anyway.

[00:03:13] Kalen: Right, right, right, right.

[00:03:15] I remember when she told me like before she did the first lesson with him, how she was gonna do it and she was like, excited and stuff. And it's like, and then she did it and she was like, oh yeah, it didn't go well. I was like, ah, that's a bummer. And then the other day they told me she's doing a lesson.

[00:03:30] And I was like, what? And they're like, yeah, they, you know, they want to keep shooting for it. Shoot, keep trying it or whatever. I was like, well, they must not completely hate, hate you. No. I'm glad she doesn't hate, hate it either, but Yeah,

[00:03:43] David: he takes a few tries Oh. To, to get she the. Yeah, and I think the second time my wife sat with him and that, and that went a little bit better, so we'll see.

[00:03:52] Kalen: Yeah, dude, that's so cool, man. That's awesome.

[00:03:56] David: Little bit of lore for everyone out there.

[00:04:00] Kalen: Did you see the tweet tell some lore about yourself and then No.

[00:04:04] David: Oh yeah. All the different words and it's,

[00:04:06] Kalen: I still don't exactly know what lore means in that context, but I kind of understand, but I was like trying to think about something to post and I was like, I know man, like everything seems stupid, so I didn't post.

[00:04:21] I feel like you have

[00:04:21] David: some lore you have like stuff that you've started up and sold and everything. I feel like that's a cool little lore. That's lore situation. Okay,

[00:04:28] Kalen: gotcha. Gotcha. I guess that makes sense. Dude, it's funny because I was actually disappointed, like when I realized we have to talk about the PIM stuff afterwards.

[00:04:38] Okay. To the audience, to our vast audience, we're gonna be. Jamming out on a Claude Pim idea. That is all. Oh yeah. That is all you get for now. But anyways, so we're telling everyone they're gonna

[00:04:51] David: beat us to it.

[00:04:53] Kalen: I know. So I was getting doing my pod prep and then I was like, oh shit, if we talk, uh, and do the PIM thing first, all the like casual ketchup material is gonna be lost.

[00:05:08] And so then I was like, we gotta do the pod first. But then I'm bummed because I'm like dying to talk about

[00:05:13] David: this Claude Pim thing. That's how you know that we care about this podcast so much. Yes. It's much We gotta capture all the value.

[00:05:22] Kalen: Yeah. We are capturing all the value. Bringing it to you guys direct.

[00:05:28] Gigantic. Shout out to Miles . Who rode in? , I'm not making that name up. It's like such a cool name that it sounds kind of borderline fake. But anyway, miles rode in and like he said, he is liked the podcast. It was like, it was cool. Oh, and then also he was talking about the robotics.

[00:05:45] 'cause he heard the episode, we were talking about robotics and he pointed me to a Lego. By the way, we have several Lego topics for you today on, on the pod, but I just connected those dots. But like a Lego robot set that cost like 400 bucks. And I was like, that's perfect because Shees, the first one I tried was like 150.

[00:06:06] Uh, at the time I was like, it's the same thing with these AI tool prices we're paying. It's like, at the time I was like, ah, that's expensive. But it's like, fine if it works. And then it didn't work. It was the, it was just cheap. The Bluetooth connection was cheap. So now I feel like 400, I feel like I can trust that price point a little bit,

[00:06:24] David: you'll have to let me know how that turns out. I feel like in my situation, that would be a, yeah dad, this is cool for like our tops and then Right, it goes back in the closet.

[00:06:36] Kalen: Well, but that's the great thing about a 30 day return. And I do that, I literally do that all the time. It'll be like, Hey, if you're not using this after two weeks, it's outta here.

[00:06:46] You know what I mean? That's what's great about that. And I returned that $150 one actually, he did start using it less and less, but it was also 'cause it was just, I don't know, there wasn't as, it would keep disconnecting a lot and then he got to reconnect the Bluetooth and stuff like that.

[00:07:00] So, oh my God, I hate Bluetooth so much. Can we get Elon to work on Bluetooth next? That would be, please let's solve

[00:07:08] David: this.

[00:07:09] Kalen: Yeah, I think super intelligence, self-driving robotic spaceships to Mars and then maybe, maybe then Bluetooth.

[00:07:19] David: That might be the marker for whether or not we've hit a GI. If we can solve the Bluetooth problem once and for all,

[00:07:29] Kalen: what is the deal?

[00:07:31] Everybody's constantly debating like, have we reached a GI? How do you know you just solved it, dude, you just literally solved it. That's it. That's all you gotta

[00:07:41] David: do.

[00:07:41] Kalen: Once they can fix Bluetooth, that proves that they're smarter than anyone else on Earth because no one else on Earth has fixed it.

[00:07:50] David: Yeah, apparently like

[00:07:51] Kalen: right,

[00:07:52] David: there was a like, do you remember the Nintendo, the switch that came out?

[00:07:56] Like they had Bluetooth issues on their controllers because one of the problems with Bluetooth is it doesn't go through the human body well. And so if your controller is small and your whole hand covers it, it just doesn't work.

[00:08:10] Kalen: Gotcha.

[00:08:11] David: By the way, can we please fix the Bluetooth

[00:08:13] Kalen: by the way? Like. That's not the worst thing in the world.

[00:08:18] I mean, I'm constantly terrified that I'm frying my brain with my air pods. I mean, everybody tells you that, you know, if you put your phone on your lap, you're gonna fry, you know, your balls and stuff like that. Mm-hmm. I mean, yeah, like maybe it's a good thing. I dunno thing that, that

[00:08:32] David: it doesn't, but also, yeah, I get a little scared when I just put my phone.

[00:08:37] If I take a nap and I put my phone on my chest, I'm like, maybe I should just put my phone over over here.

[00:08:42] Kalen: I know I'm gonna have a heart attack. We're all like terrified of fucking everything these days. No, but you know, it'd be funny. It's like, you know how sometimes like you tell a developer like, oh, there's this bug, you know, whatever.

[00:08:56] And then, you know, can you fix it? And then they fix it, and then they cause like a much worse problem because , they just sort of carelessly, you know what I mean? Did it without like really thinking it through. Like imagine if the Bluetooth developers did that. They're like, oh yeah, we can fix it.

[00:09:12] Whoop, we can make it go through your body. No problem. And they fix it. It's like a year later, like it turns out it microwaves you. 80% of the population gets cancer like overnight.

[00:09:24] David: Um, maybe we shouldn't give that problem to ai.

[00:09:28] Kalen: Okay. So many interesting topics. You have a decent list. Your list, by the way.

[00:09:35] My list, and I have pictures, I have receipts. David didn't believe me that I had a list. If you go back to 10 episodes in Totally, and I would talk about the list, which is on my computer, and I didn't wanna share because I didn't want him being distracted by the topics, just so that they'd be organic. And then you said at one point, I, I don't even think this list is real.

[00:09:57] So now I write it out because also that's my new flows. I don't look at the screen at all and I send it. So we have the receipts. My lists are legit. You have finally gave three items on your list today, which is,

[00:10:10] David: I think it was two items.

[00:10:12] Kalen: No, no, no. It was three. You said these two and then you said, and something else.

[00:10:18] Oh, okay. Well, I already forgot what's on my list.

[00:10:24] Okay. All right. I maybe I shouldn't have complimented your listing. Yeah,

[00:10:29] David: no, just,

[00:10:29] Kalen: yeah, that was a little immature. Okay, so I've been messing around with Claude headless dangerously dash dangerously accept permissions.

[00:10:39] David: Oh my god, I'm so scared of that.

[00:10:41] Kalen: Yeah. Okay. So the shop worker thing is an open source, basically CLI tool to create web hooks , and deploy code to CloudFlare and.

[00:10:50] Shout out to open source. Shout out to open source. And, but I was like, well I gotta create an app for it. And so I spun a little app and it has like docker containers that run clot headless in them and stuff like that. Right. It's funny 'cause I was thinking about the open source thing as I was like, I'm starting to feel a monetization vibe again, which will probably only last for like a couple weeks and then I'll give up on it for the ninth time in a row.

[00:11:16] But I'm starting to be like, okay, I have a little app idea, I'm gonna run with this for a little bit time to monetize some shit. And I was like, I wonder if I should keep it open source. And I, part of me was like, no, let me not do that. I was like, yeah, I need to lock this shit down so I can start charging for it.

[00:11:31] But then I was like, , screw it. Like I'm gonna keep the core thing open source and then I'm gonna just throw this app on top of it. Who cares if it works or not, whatever. But it basically like, it clones the repository in. To the container and then it like basically sends CLI commands to like run the CLI stuff you would run locally, like to deploy or to test a job or whatever.

[00:11:55] Yeah. Like the way that I personally like to work on this type of stuff is in my local environment. If I have something I gotta do for a client, I'd rather do it locally and test it locally or whatever. So that's how it works. Like the app wraps all that stuff, but then if you wanted to just pull it down to your local, you pull all the source code down and test it and stuff like that.

[00:12:16] So it's kind of like a neat little model.

[00:12:19] David: How does the claw and headless fit in there? If you've got the repo of the shop worker

[00:12:25] Kalen: stuff? Yes. Yeah, so you have the repo of the shop worker and basically the idea is that somebody can just install the app and then do a prompt. And say, Hey, when an order's created, if the line item has a custom property birthday, then tag the order as whatever and update a meta field and send an email to customer service.

[00:12:52] Boom. So then what it does is that it takes that prompt, it runs clot against that code base. That's why I, I think that's why that was the only way to do it. Like I couldn't just do like a clot, API call or something like that because it has to like. Work against the code base. 'cause the code base has like patterns for how to create a trigger, how the code is supposed to be structured.

[00:13:20] The code is structured into like steps so that it can deploy to CloudFlare. And Claude is really good at looking at your existing code base along with like a Claude MD and figuring shit outing. So basically That's interesting. Yeah. So, so the

[00:13:36] David: repo is just like a set of tools that you gave Claude to work with locally?

[00:13:41] Kalen: Yeah, well it's a, and the, yeah, it's like a set of tools and code patterns, because like you could just use Claude and just say, Hey, create a web hook and write some code to do something and it might work, but then like it wouldn't know. For example, what if you said, okay, loop over 10,000 products, then it would most likely just go.

[00:14:04] Okay, loop over 10,000 products. But then in the CloudFlare worker environment or in whatever environment, you might hit a request timeout, you might run outta memory, right? So it needs to have like a framework in place for like, that's why the CloudFlare workflows are cool 'cause there's a framework for splitting up work into steps so that it can like scale and stuff like that.

[00:14:30] So there's like enough structure in place for it to do this specific thing, but it's like it's pretty light, it's like a very light structure's.

[00:14:40] David: So lemme see if I understand this. This workflow, a merchant expresses something that they would like to have happen on order create, and every time an order is created.

[00:14:52] You send a web hook to this worker that has Claude installed, and Claude interprets the,

[00:14:57] Kalen: so the worker doesn't have Claude installed, the worker just has the deployed JavaScript code and the web hook, URL and stuff like that. So the Shopify app spins up a Docker container, which is a separate instance.

[00:15:14] The Docker container then deploys to CloudFlare via the Wrangler command line tool. Yeah. When they're prompting in the app, it changes the Docker container, and then Claude is running against that copy of the source code there, and then we wanna deploy it. You deploy it to CloudFlare, and then there's no Claude on the CloudFlare side.

[00:15:37] It's just deployed code.

[00:15:39] David: Interesting. So it's not like Claude is triggered on every one of those new orders? No, it's, no,

[00:15:44] Kalen: no, no, no, no, no. It's Claude interprets

[00:15:46] David: what you want and then builds code to do that thing.

[00:15:49] Kalen: Yeah, it's, it just like at design time you're using Claude. Cool. Yeah. And then you could also just, use it locally.

[00:15:57] Like just you write the code without Claude if you want to type of thing. But anyways, so the Claude headless thing is kind of cool. And I know we're gonna talk about like maybe some CLO stuff with the pim, but it's pretty cool to be able to run Claude headless. It's also like a little terrifying. And I'm like thinking, is this an absolutely horrible idea?

[00:16:18] I mean, I don't know how much I did check to make sure that, for example, my Claude Key that it gets tied into Docker is not accessible via the Claude prompt that runs because mm-hmm. At first I didn't have that set up right. And then I, I was like, I realized, oh, let me ask Claude for my. API key and it gave it to me.

[00:16:43] So then I was like, oh shit. So I had like, you have to put it in another directory so it's sandboxed out and then Claude says, I can't read that file. So there's some things like that. But other than that, I don't know their API token is there, but it's, just their own API token for Shopify. So it, just has the permissions that they already have.

[00:17:04] David: So

[00:17:04] Kalen: I'm

[00:17:04] David: not, so when you, sure. Like when someone installs this app, does Claude handle grabbing the API key or you have to give it to 'em somehow?

[00:17:12] Kalen: You give it to 'em. So you, when you spin up the Docker container, you spin up the Docker, which is like a little Linux machine, and then you cannot run it in root mode, which I tried to do.

[00:17:23] 'cause I, I had didn't create another, I was just, Hey Docker, make a thing. Okay, do this. And it was doing it under root. So then you have to create another user. You gotta go into that user to use Claude and, which is cool because now then Claude doesn't have access to you. I later realized why that was a good thing.

[00:17:42] David: So does that mean you have some setup to do? Like when someone installs this app, you also need to. Spin up a Yeah. Docker container for them.

[00:17:50] Kalen: Gotcha. Yeah. Which was wild because I've been here about Docker forever. I remember like when like Mark Schutt in the Magento community, like Yeah. Did a docker thing for Magento and like everybody was hyped about it and stuff like that.

[00:18:05] I kind of want to understand it, but I just don't like, it just felt like the type of thing that I would've definitely spent a couple days screwing around with. Hit some annoying error, spend way too long, browsing Stack Overflow and then just, , give up on,

[00:18:23] David: that's definitely how it was.

[00:18:24] Like we tried to make our own like Magento go kind of thing.

[00:18:28] Kalen: Okay.

[00:18:28] David: For a little while I remember. Yes. Nucleus, right? No, that was Classi Lama. That was after I left there, but we were trying to like. We had a client who wanted to be able to spin up Magento instances, like within seconds, like someone signs up and then boom, there's a Magento instance.

[00:18:46] And we used Doctor for that. Was this say

[00:18:47] Kalen: AOE or, yeah. Okay. Dude, I just remembered that you worked at aoe, like only ever thought about Classy Lama for some reason, but that was such like an elite team dude, with like Fabricio and stuff. Oh my God. It was

[00:19:01] David: awesome. That's, yeah, I learned so much working.

[00:19:03] There's, so, yeah, that was awesome. I, I worked there for a long time and learned a ton of things about like AWS and Nice, like Docker for a little bit. It was crazy. Nice. But yeah, we tried to do that and there, there were so many difficulties with trying to get, like everything spun up to run Magento in a Docker container.

[00:19:24] Okay. And like we have these like Apaches on its own thing, and the MySQL is installed in its own container and they had to communicate. We probably made it way too complicated. Right, right. I wonder. You could probably have it spin up a Docker container like on app install so that you don't have to do anything manually.

[00:19:42] Right,

[00:19:43] Kalen: exactly. Yeah. Yeah, that's, so it's currently there. Well, there's like a button I click in the app and it does it, but then I'm gonna tie that into Oh, gotcha. Then I'm gonna tie that into automatically upon app install. The only, the reason why I don't do that first off is because it's so hard to test.

[00:20:00] Like you're like, I'm not gonna uninstall and reinstall the app 19 times as I this function. So then you go, let me put it somewhere else. You put it somewhere else and then you don't get back to moving it back to app installation until. The very last minute when like, that makes sense.

[00:20:17] David: But yeah, that's cool.

[00:20:18] You got some cool stuff. Cooking,

[00:20:20] Kalen: dude it's funny you confirmed my, what I thought about Docker, 'cause I just remember that was my sense that it was just super complicated and I'm sure it's gotten easier. Like, I don't know, I'm guessing there wasn't like Apache setup and also my setup is probably way simpler than yours.

[00:20:35] Just with like doing stuff with NPM and whatever, it just kinda like mm-hmm. It just kind of mostly works and Yeah, like cloud code just banged. I mean it, it definitely, it did hit some snags and stuff like that and , just like typical, typical stuff, you know. But then what was funny was. So for the first day or two, I was on a tear, dude.

[00:20:58] Like I was ripping through the Polaris like so fast and like, yeah, the docker thing just worked. And then I got like the clo headless thing and then when I was trying to get the logs to stream into the app, I like got stuck on it and I was like, vibe coding it so hard. That like I wasn't reading barely any of the codes.

[00:21:23] And then I was like, this is incredible. It's going so fast. And then I like hit a wall and it stopped. It couldn't fix the streaming. And I was like, it's not working, it's not working, it's not working. And I was like, shit, am I on Opus? And I was like, yeah, I'm, I'm on Opus. And it was just stuck. So then I was like, man, this is a stupid idea.

[00:21:46] And then I kind of was about to give up on it. I was like, this architecture makes no sense. And I'm like, and it's an interesting effect of like, when. The vibe coating is going really well, and it's like really easy because it's just doing really well and you've, you don't have to think that much when it stops and it inevitably does, when it hits a wall with something, it's really hard to get over that wall because of like how easy it was going.

[00:22:19] And then all, like, the difficulty curve goes from like 1% immediately to 90%, and I think it's as much as yeah, people can do loss of vibe, coding, whatever, people that don't even know how to code, like I, that wall is brutal when you hit it and you gotta be willing to sit there and figure some stuff out.

[00:22:41] David: Definitely. I like when I'm in that mode and I'm doing something and it feels like. Claude's hitting all the things that I wanted. I'm like, I know this isn't gonna last forever. Right. How do I, like, I have like context anxiety. Like this context right now is perfect. Right. And I know it's not gonna be that way, way for long anxiety.

[00:22:59] I gotta, like make sure that my take advantage, this new prompt that I add is like not gonna mess something up or if I feel like I'm going in a different direction to just start over somehow. Right, right.

[00:23:09] Kalen: No, and, and you really do have to just like refactor, like if your code, you have to be able to look at it and for it to make sense.

[00:23:18] Like even if you're looking at it pretty, glancing at it pretty quickly, you know, if it doesn't and then the architecture gets like, oh, and that's the other thing I is that I heard about cloud code agents and I used, I set one up. It's weird, like you type slash agents and then it says you can create a separate agent to do specialized things like a front end engineer or a DevOps or security reviewer or whatever.

[00:23:46] And I'm like, okay. And it listed like code cleanup specialist. I'm like, okay. So then you go next and it says, okay, create your agent, and then you just start typing. So then I was like, code cleanup specialist, enter. That was all I typed. It wasn't like a selection field, it was like a text entry to define your agent or whatever.

[00:24:11] I'm like, okay. So I do that and then in order to use it, you just say in the prompt, you just say, use the code cleanup agent to like clean up this file. That's it. And I swear, I feel like the refactoring is like. A hundred percent better than it was before. I just had it ripped through a couple files and it renamed the method.

[00:24:36] Names were just really clear. Nice separation of concerns,

[00:24:42] David: perfectly Campbell case.

[00:24:43] Kalen: Yeah, I scanned it to make sure everything made sense, you know? I was like, yeah. Yep, yep, yep, yep, yep. But it was just a lot cleaner than I remember a regular refactor being even in cloud code Opus,

[00:24:58] David: although I did.

[00:24:58] So is that the use case? For subagents? Is it just to keep context clean in the top agent or what? Like why? Why

[00:25:05] Kalen: do you do that? So here, here's the confusing thing is that I don't think it's a context thing like. I was in a chat session with the main chat and I said, tell the code cleanup agent to do this.

[00:25:17] And then it says, cool. And then it shows a little blue label that says code cleanup next to whatever, some of the actions that it takes or whatever. Okay. But I think the context is the same. I think it's just a different system prompt or something for Gotcha. You know what I mean? For the thing. Yeah. It seems to make a significant difference.

[00:25:37] Like I'm pretty sure too, the next thing I'm gonna do is like the security thing seems like a no brainer is like, and I think they added a specific slash command for security reviews. I'm sure that'll be significantly better than like generic or if you don't even ask for a security review.

[00:25:53] If you just tell it to do something and it does something according to standard practice.

[00:25:58] David: So many things I don't understand with all of this stuff. Right, right. I seem to keep going with it.

[00:26:04] Kalen: There's so much. Yeah, there's so much mystery.

[00:26:07] David: I've been trying to watch those videos that Anthropic keeps releasing on, like cloud code best practices and stuff.

[00:26:12] They're doing a great job of pumping out content. Oh,

[00:26:15] Kalen: nice. I'm not sure I've seen too many of those.

[00:26:19] David: I gotta, there's a few of 'em. They have 'em on YouTube a lot.

[00:26:22] Kalen: Dude. They're such an impressive company. It's unreal, dude. I mean, it's open AI obviously is insane, but , the fact that they're just,

[00:26:31] David: did you see the guy that made Claude Code?

[00:26:34] Boris, I think is his name. Hopefully I got that right. Mm-hmm. I don't know if that's his name, but he got poached by a cursor and then mm-hmm. Like a week later he was back at Anthropic. Kind of an interesting trade.

[00:26:49] Kalen: That's funny. Wait, did Anthropic end up acquiring Cursor?

[00:26:57] David: No, I don't think so. Oh, okay. , There was something like maybe someone was gonna acquire cursor or someone did acquire Windsurf.

[00:27:05] Windsurf like halfway. I don't know. That was, yeah. There's always drama going on in this ai I know realm. It's kind of fun. I know

[00:27:13] Kalen: there is a very fun, like, drama factor to it.

[00:27:18] David: Yeah. I was watching the GPT five release yesterday for like five minutes and then I was like, it was a bunch of graphs and the graphs were all weird.

[00:27:28] And so I was like, oh, I'll look at this later. Everyone has been complaining about the graphs. The graphs, what is going on. Yeah. 'cause they had like a bunch of bar charts that, , what is it, like the height of the bars just didn't make sense. Like they were trying to play with how it looked. Okay, gotcha.

[00:27:47] And it was just, it was weird. I, I got a weird feeling about it. Gotcha.

[00:27:53] Kalen: Yeah, I didn't check that out. I thought the demo looked cool, , and I think all the demos were live. If I recall, I asked it a few questions. Like I specifically went back and asked it a couple questions that oh three got wrong and it, or that I thought it oh three got wrong and it got 'em right.

[00:28:13] Like it's, what

[00:28:14] David: kind of

[00:28:14] Kalen: questions do you ask? Okay, so one was when I broke my finger, I had the x-rays. I uploaded the x-rays to four oh and oh three and also gr. And I was like, is the finger broken? And what would be the treatment plan? And they all basically identified like the angle of the break, which was kind of crazy.

[00:28:35] And it was like under a threshold where you wouldn't require surgery. And then, but then they all said to wear a splint. And when I went to the urgent care, they gave a splint. They gave me a splint, which was basically like as big as the cast. And then when I went to the hand specialist, he said they don't actually recommend that anymore.

[00:28:57] They recommend these Velcro loops either. It's called a buddy loop, it costs like a couple bucks. It's like a really simple thing, but it's actually better. Basically

[00:29:06] David: a piece of tape. That's what they used to do. Yeah. When you got your hand stepped on in football. Yeah, just tape your fingers together. Yeah.

[00:29:12] Kalen: And because he said the slings make your hands stiff, they're actually not as good for recovery Anyways, that was the detail that GT five got. Right? Is it mentioned the buddy loops and then that's so interesting. Um, yeah. And then the other question I asked was about using CloudFlare. It was like something with compression and it got it right.

[00:29:32] Like and the other ones like messed it up.

[00:29:35] David: I feel like the health related questions are like the secret. Well, I mean everybody knows, but . Such a helpful, like my wife uses that all the time, dude, it's huge. Just to ask like health questions,

[00:29:46] Kalen: it's huge. Yeah. Yeah. It's so cool. I got a, the community, Lego, the Lego guy.

[00:29:56] David: That's awesome. Did Shopify send that to you?

[00:29:58] Kalen: Yeah, it was, it was like for posting in the forum, like if you had over 150 posts in the forums or something like that. Dang dude. Yeah, dude. So that, was, that was pretty pretty. I say, you know, you're an absolute nerd when you're excited about a little Lego thing.

[00:30:14] Just like a complete nerd.

[00:30:17] David: I love it and I love that Shopify takes care of people. Like do does something for people who are. Post stuff in there. Yeah. By the way, I just realized that your notes are written on the back of what looks like a multiplication table is correct. Homework assignment. That

[00:30:33] Kalen: is correct.

[00:30:36] That

[00:30:36] David: is correct. Oh man, I have flashbacks from like fourth grade having to go through those as fast as possible. That's so funny. You're

[00:30:45] Kalen: just staring at it. You probably were able to tell by zooming in on it,

[00:30:49] David: right? I, I was just looking at it. I was like, why are all these boxes on here? Must be some interesting like grid paper that Kaylyn's got.

[00:30:57] Nope. Just stole his daughter's homework and wrote notes on the back.

[00:31:02] Kalen: Just lifted her homework. She doesn't

[00:31:05] David: need

[00:31:05] Kalen: this. That's so funny. So I saw that Yapo shut down their, , email and SSS apps.

[00:31:15] David: Yeah, I wasn't like, I, I kind of missed that I saw something about things happening and then my boss was asking me if I saw it.

[00:31:21] 'cause like when we were trying to sign up with YoPo or renew like something with them, they kept trying to throw in those like email and SMS things and we're like, nah. And I guess they, I guess it was a lot of people that said nah to that. Huh.

[00:31:37] Kalen: Gotcha. Yeah. That must be it. Yeah. It's, funny, I, the first post I saw was like something, the YoPo thing and I was like, what's the YoPo thing?

[00:31:47] What is it? So I had to, I asked Rock actually, I think I asked. Oh. Like I'm using that a lot dude. 'cause like there's so many times you see a tweet and you kind of don't understand it and you kind of want to, but you're just like, ah,

[00:32:00] David: who cares?

[00:32:01] Kalen: And it

[00:32:02] David: solves Iactually do appreciate. Yeah. Like when people post, it solves for that

[00:32:06] Kalen: perfectly.

[00:32:07] It

[00:32:08] David: does. What is actually happening in this thread right now? 'cause I don't have. Any idea? Yeah,

[00:32:14] Kalen: because like so many things on Twitter are like, there's not enough context, which I like, because if you follow people closely, it's better that they don't add context, you know what I mean? Because it's like this ongoing conversation.

[00:32:27] So it's better that there's less context and you're kind of on the, in of the context, but whenever you're on the out of the context, it's so aggravating. It's like just say it, like spell it out. It's a really funny thing.

[00:32:44] David: It's bad for engagement if you give everything away in the first tweet.

[00:32:48] Kalen: Right, right, right. Yeah. I saw some random Twitter account yesterday. Do you ever do this? Where you scroll? Dude, I feel like an absolute stalker, but I just scrolled through like their tweets for like a solid like 25 minutes. Like have

[00:33:08] David: you ever, yeah, sometimes somebody will pop up that I haven't seen before, and I'm like, who is this?

[00:33:12] Yeah. Oh, Kaylin follows them. Okay. They must be saying something and then it's just like a Shopify themed shit poster, so yep, I'm following that

[00:33:24] Kalen: show title, but sometimes I'll spend like an inordinate amount of time doing it. And I remember before I used to like, like multiple tweets. Yeah. So I'll give them like a power, like, you know,, and then, because I genuinely, like, I've just read 97 of your tweets and like genuinely these four are like really funny.

[00:33:49] But then I'm like, they're gonna think I'm a psycho. Like so now I don't like anything, you know? I just like let it be. Yeah. It's very weird. But you, with the yapo thing, it's kind of weird, man, because so many people are getting so much more done now with AI that it's kind of surprising, I guess I couldn't help but somehow imagine it was tied to AI in the sense that maybe it's tied to the kind of the changing business dynamics of how companies should be structured or what, like I figured it had to somehow be related to everything that's going on in the industry.

[00:34:26] But maybe it was just like you're saying Rev, maybe it was just a revenue thing just happen. They happen to decide now that just No. Yeah. You know, just is not converting. I don't know.

[00:34:37] David: I think a lot of like SaaS, like big SaaS players try for some reason to just keep expanding their product set.

[00:34:45] Kalen: Right.

[00:34:45] David: And like that also isn't a great thing for focus, like from a company perspective.

[00:34:50] So maybe this is a good change for them and they can focus more on their like core products and ends up being good overall. But yeah, like that was a thing for a long time. All of those guys were just like, got super big on like reviews or something and then kept reaching out to compete with like way too many other products.

[00:35:11] Kalen: Right, right. Yeah, no, I mean I totally get that. And I don't like, generally speaking, I don't like companies with a bunch of apps, but it does seem to be a successful strategy. I would assume that more people are going to be doing more apps. Like, do you follow the Revo guy, Stewart? No. Maybe, dude, you gotta follow him.

[00:35:35] Dude, he's, maybe I should, he's like the most advanced AI dude, I would say in Shopify tweet. I need that. He like his company, like they use Devon. They've used Devon. They have it like heavily instrumented to where it automatically runs unit tests, does like integration te automatic integration tests, like the whole thing for their SaaS app.

[00:36:02] And, um, found

[00:36:03] David: him

[00:36:04] Kalen: following. And he's always talking about how like they're seeing like 20 to 40 x gains or whatever. And I've been following along and I'm like thinking to myself like. What are they going to, are they gonna just add a ton more features to their loyalty app? That almost seems insane. And then he ended up posting that they're, they built a new app and he is , we got two years of like engineering done in like 90 days on like an app that's in like a category that's kinda like a VC-backed category.

[00:36:36] In other words, it's like, there's shitty options and I, I can't wait to see what it is, but I feel like more companies are gonna do more apps is kind of my instinct.

[00:36:48] David: I mean, at least the people who are really ingraining AI into their workflows are gonna be ahead of everyone else on productivity.

[00:36:57] Yeah. I just wonder like you still have. Maybe he's way more advanced and has all these different agents who are focusing primarily on product and like understanding how to improve things. But that just seems like a way to overload yourself. But if he solved that overloading problem with AI, then that's awesome.

[00:37:16] Kalen: Yeah, that's what I would assume. I mean, it's gonna, all the productivity gains are gonna have to go somewhere. Like we're either gonna all be using like much more feature rich applications and I don't even know what that looks like. Like I can't even imagine give me twice as many features in like this app that I like and how could that be a good thing?

[00:37:38] David: Yeah. There's also that thing that kind of scares me, which is like when something is doing well and it ends up getting a bunch more features because people requested it, it starts to get. Overloaded and like crufty and things don't fit together well. So

[00:37:53] Kalen: yeah,

[00:37:54] David: if it also helps solve that where like you still stay smart with your base architecture of what your thing can do and also support everyone's use cases.

[00:38:02] Kalen: Yeah.

[00:38:02] David: That's cool. Its just not, that's the, it's not typically been like that.

[00:38:06] Kalen: Yeah, yeah, exactly. But it's if you have a solid app and all of a sudden you have three x let's just say only three X productivity or five x, maybe it's 20 x, whatever the real number is, like where is that gonna go,

[00:38:21] or I don't know. It'll be interesting to see if , 'cause it's true, like most of the like app companies that like you like that really have done a good job over time. They focus on like one app more or less. Mm-hmm. And the app factories that have, , I mean there's some companies with a couple, two, three solid apps, but I don't know, I'm just saying like.

[00:38:45] I think generally it's like a quality signal when there's like a company that has one app type of thing.

[00:38:52] David: Yeah. And it's, I never feel like it's because they have some miss on the number of engineers that they have. If you're just thinking about the amount of code you can generate, generating more codes, probably not the best thing to do.

[00:39:06] Right. So they must have something like an AI agent, product manager, developer kind of thing that can look at metrics and understand what's actually important versus just writing a ton of code.

[00:39:20] Kalen: Right, but then that, limits because can, when you can have developers building features, that's where you can create value very quickly.

[00:39:29] So, I mean, which is what a SaaS company's is trying to do. So like if all of a sudden. You have this resource that you can't deploy at the same rate, then you're either gonna just wanna fire them or you're gonna want to do more stuff. I don't see a company with three developers that's doing well, wanting to fire two developers just because they can.

[00:39:54] I feel like they'd wanna do something, like they'd wanna put that into something else as a team. You know what I mean?

[00:40:02] David: True. Maybe that's like more code review or something, or, and then you have Gavin.

[00:40:07] Kalen: They're like, I don't know. It's weird. Let's all spend eight hours of our week reviewing each other's code.

[00:40:13] There you go. It would be like code. So reviewed. It'll be like the most reviewed code ever. They're like, we're gonna, this is gonna put that in market, this awesome market. I can't wait. We're gonna make 9,000 mermaid diagrams. Everybody's psyched.

[00:40:32] David: Oh, we could definitely use more diagrams in the world. I love a good diagram.

[00:40:37] Kalen: Dude. Yeah, I, that's a little flow

[00:40:39] David: chart,

[00:40:39] Kalen: right? I'm thinking about my buddy Colby who usually listens to the podcast and he's a gigantic chart guy. Yeah, he creates flow chart and I'm generally anti flow chart, so like I was always making fun of him about it in earlier projects we worked on and like it would be outta sync 'cause it was like this big ass flow chart and then like the details obviously changed and then I was like, oh, it's an outta date, I'd make fun of him and stuff like that.

[00:41:11] And then we worked on a couple other projects and he did flowcharts too. And then at some point I was like, I kind of respect the flowchart. Like the first, the very first one he did was, was too much. But now the level he's doing in that, I'm like, I see the value in this. As much as initially I was like, yeah, just

[00:41:30] David: silly.

[00:41:31] I respect the flowchart game. Like, yeah. People, like developers I think tend to maybe visualize things internally, like in their mind and it's hard to communicate, right? Without having a flowchart. Like I spent maybe five to 10 minutes making a, like when I first joined this company and people still ask me for it.

[00:41:51] Like, it's like it's the best way to convey certain things, I think. 'cause people are really visual.

[00:41:58] Kalen: Yeah. I think people create a flow chart and then like you get that feedback somehow that it was like incredibly valuable. And then you're just sold for life.

[00:42:09] David: They're just flowcharts bang them out from now on.

[00:42:13] Kalen: I think, one positive flowchart experience in life is enough to like be committed to flowcharts for the rest of your life. Even if it was a gigantic waste of time and you really didn't like it, you just would remember how good that one flowchart experience was.

[00:42:33] David: I'm gonna make a Claude Subagent that always asks if I want a flow chart before it does anything, and I will.

[00:42:41] That's an auto complete. You know what would be, just always start with the flowchart club.

[00:42:45] Kalen: It'll be really interesting to see if AI companies do like April Fools jokes. Because that would be an awesome April Fools joke.

[00:42:57] You know how like low chart Claude, like April Fools jokes used to be like super fun and like you'd look forward to them and then all of a sudden it was just like, this is dumb. Like I'm over it. Like these are such like rehashed ideas this would make April Fool's jokes great again, dude.

[00:43:15] David: There you go. Yeah. Just totally ruin everyone's productivity for an entire day with something I, want them to bring back Golden Gate Claude. That's my favorite. What's that? That was like Anthropic was testing out, like what happens if they change certain things within the model. Mm-hmm. And like trying to understand what they all related to.

[00:43:38] 'cause you know, it's like a, like crazy dimensions of things going everywhere with matrix multiplication. Mm-hmm. And so they were trying to understand that a bit more and at some point. They accidentally made Claude think that it was the Golden Gate Bridge and every time you prompted it, it would like relate it to the Golden Gate Bridge.

[00:43:56] That's so funny. You could ask it like what the migratory behavior of ants was and it would be like when ants are trying to cross the Golden Gate bridge,

[00:44:06] Kalen: that's, have we never talked about that? That's, I feel like I tell everyone about Golden Gate product, it's my favorite thing. But dude, that would be an incredible April fool.

[00:44:16] Like, and not as an opt-in, like as an automatic, like a automatically it does something kind of goofy, but like it doesn't derail your work. 'cause like people are depending on it for work. So like it can't really derail you for more than a few minutes. But like yeah,

[00:44:32] David: it's gotta be the equivalent of like your coworker showing up in a clown costume, but they're still work done.

[00:44:37] Kalen: A hundred percent. And if it could do that to everybody in the world. Like at the same time, like that would be incredible.

[00:44:48] David: Oh, I'm looking forward to it. And now we gotta manifest this somehow.

[00:44:51] Kalen: There's, yeah, we've got a whole manifestation list. I gotta fix our damn that because that's funny because that would be a fun thing to search if we've said that.

[00:45:01] I don't know, a few different times. And a couple times I've done that with the AI search thing on the site and like it kind of worked, but then it kind of like it missed stuff I knew it shouldn't have missed. That is a really fun like AI search use case. I gotta try to tie it into a different vector database or something like that.

[00:45:21] But that's a fun little use case to be able to search for stuff.

[00:45:26] David: Yeah. That thing's still working. Right? You could build your own little system. Prompt injection in there for April Fools too.

[00:45:32] Kalen: Yeah. Yeah. Oh yeah, yeah, yeah, yeah, that's true. That'd be funny.

[00:45:38] David: I was looking at the MCP storefront, MCP stuff. Oh, nice.

[00:45:42] Again, and it seems like they're, it seems like Shopify's working on support for cloud desktop, but I still feel like it would be interesting to, and then I, they came out with those tweets about agent commerce being a thing.

[00:45:56] Kalen: Right,

[00:45:57] David: right. And I'm still anxious about being able to like, optimize what it is that the agents are getting from my store.

[00:46:06] Oh, that's interesting. Because like, you can't do, you can't make any, like we have variants that we don't wanna sell. Right. But they just all come through right in the MCP. So like, how do I change that? And I was thinking like, maybe I can just inject some hidden HTML into the description to get the agent to do something.

[00:46:25] But hopefully they're gonna gimme some tools there.

[00:46:29] Kalen: Yeah.

[00:46:30] David: Yeah.

[00:46:31] Kalen: That's cool. I saw Toby tweeted about it. I was like, really trying to understand that the idea, like that it renders little product cards. Mm-hmm. In chat GPT or something like that. Like

[00:46:45] David: I don't Yeah. With the remote dom.

[00:46:47] Kalen: I Just don't quite understand the whole use case.

[00:46:51] Like, it's literally just about displaying a nice little HTML widget in chat GBT, that's basically the idea. Yeah, I think so. Okay. Yeah, I mean, that's cool, but like, I don't feel like that's world changing. I mean, it's like you can click on links now. I mean, so it, it adds an image on top of the link and a price, like, okay, cool.

[00:47:12] Yeah. Is that gonna change everything or.

[00:47:15] David: I bet. What if the image is a gallery and so the customer's like actually doing some shopping within the chat window and they can make decision,, you're not gonna make a decision based on how a link looks, but you might make a decision based on the silhouette of the product or like some sort of information.

[00:47:31] Kalen: And is it gonna be super early? Is it gonna be an actual rendered web like component? I know it's a web component, but it's an actual thing where you could have like whatever, you could have a product gallery in there and add to cart button, like sessions. Like, , is that how it's gonna work?

[00:47:50] David: That's how I imagine it.

[00:47:51] Like you, because even with the storefront MCP, it creates a, a cart session for you and it, it can modify that cart over time and then send you to checkout.

[00:48:00] Kalen: Okay. That is pretty wild. That is pretty wild.

[00:48:04] David: Interesting. And like the remote MCP has like a little JavaScript payload that Shopify has that I guess provides more.

[00:48:10] Right. Here's how you would switch between images, for example.

[00:48:13] Kalen: I really have been wondering how things are gonna, I, I impact the web because so much of the time I, I don't wanna see other websites. Like I just want the answer, and I've been thinking like, I really like, I think I said this, I was like, I could imagine OpenAI taking out Shopify in the sense that if I could put a credit card in chat GBT and then just buy stuff without ever going to the website.

[00:48:39] Like I might go to the website to like look at pictures, but it's more like a landing page type of thing. I don't want to have to check out. I want it to know my shipping address. Yeah. Sometimes I don't even wanna see the exact price. Like if it's in a range, I'd rather literally not even see the price.

[00:48:57] Sometimes

[00:48:58] David: I think that's like exactly how this fits in. 'cause there's so much complexity around getting the order into the person that you want to purchase it from and handling all of that on the backend. I don't think open AI wants to deal with all of that and even like, yeah. Meta is canceling their checkout product.

[00:49:15] Right. Did you see that? Like they're, I did, yeah. They're just gonna use Shopify. So like there's obviously some complexity there that you probably want to, , that happens somewhere else, right? , Yeah, Shopify's payment processing is through Stripe. You never know that. But they, like Shopify does not wanna do payment processing, right?

[00:49:32] They just have a deep integration with Stripe and boom, it works. And that a hundred percent, yeah. That happens again at the next layer for commerce. Yeah,

[00:49:39] Kalen: I did come around on that. I do think that makes sense, but as far as , do I want to go visit a website and check out , as a front end experience?

[00:49:49] Like not really, but I can see how having a nice little embeddable widget inside a chat GPT could, it's weird. It's almost like. The chat interface might become the new canvas for the web.

[00:50:05] David: Yeah, exactly. Like I want to go back to school shopping. I need pencils, I need shoes in size 10. I need a red folder and a backpack.

[00:50:14] Those don't have to come from the same place. I'm just like, Claude's my boy. He is gonna take care of my shopping trip. We're gonna get everything into some cart thing just within this chat window still. And then I say, do I have everything from my list? And then I hit checkout and the right it, it kind of conveys all those orders out to the right, whoever.

[00:50:34] And maybe

[00:50:35] Kalen: you, maybe you kind of want to like some of the products you don't even need to see a picture of. And it would be better to not have to even approve Oh yeah. The picture. But then some products, maybe you want to glance at the picture real quick. And then you can see a thumbnail, maybe you can quickly click to like very quickly expand the image without totally linking you out to another site.

[00:50:57] So I think I can start to see the idea there.

[00:51:00] David: The weird thing is that someone pointed out is if you're buying five different things from five different companies, you're paying for shipping five times. So there's some interesting stuff there. Mm-hmm. But I don't think it changes the fact that this is a, gonna change the way people shop, I think.

[00:51:16] Kalen: Yeah. And I mean, they'll probably start to, they could roll out incentives to be like, if you buy from multiple places, the shipping is somehow combined

[00:51:26] David: or something free. Like a little, like if the orders coming in through this

[00:51:30] Kalen: thing. Yeah. Shopify could cover some of those costs or like they could make it a promotional feature or something like that.

[00:51:37] David: I dunno. True. Yeah. That would be cool. At the Shopify level.

[00:51:40] Kalen: Yeah. Dude, one of the things I'm trying to figure out, this is like my biggest problem right now is how do I parallel Claude? Like I need to know too. It's like crazy because even though you're literally getting something done in an hour that used to take a full day, but you, when you're waiting for 60 seconds, 90 seconds, you gotta figure out how to fill that time because like the time thinking and writing the prompt was like highly productive.

[00:52:15] And then the next few minutes, if you're sitting like, I don't know, I literally can't just sit still. I have to either start thinking about something else, check Twitter and then checking Twitter for like nine, like it's a weird workflow. It's like I need to optimize it.

[00:52:33] David: Yeah, my compulsion is like, okay, Claude's working on something. I gotta go check Slack. Gotta check email, jump on Twitter. Yeah. Okay. Now it's been 30 seconds. Can I go back to Claude and see what's going on? And with Claude code, it like gives you the little ding, so that's perfect. Um,

[00:52:48] Kalen: dude, the thing is, I cannot get the stupid ding inside a cursor, which I still mostly like to use com code inside a cursor and it's killing me.

[00:52:58] Dude. I need a freaking ding.

[00:53:01] David: Give me the ding. I need more. I have, try that more ding. Yeah. Have you tried it in VS code? I wonder if it's like a cursor specific problem.

[00:53:10] Kalen: No, I think it, I don't even know if the cloud code plugin works in vs code because initially that's what I wanted to do, but I think it be Oh, weird because I think it maybe relies on the diff function.

[00:53:23] I don't know. It's such a weird

[00:53:25] David: Have you tried open code yet? I haven't tried that yet, no. What's that? This guy on Twitter, Dax, something like this is just what he works on is like open source tools and this and open code is just an open source version of the like Claude CLI kind of thing. But you can use whatever model you want and they spend a ton of time like working with BUN to figure out like the best way to write terminal based applications.

[00:53:53] Mm-hmm. It's really interesting watching him work on this and like all the problems that he works through in terms of like making applications in the CLI, what's the best way to do notifications and they're doing like really cool CLI specific things and like pushing that stuff forward.

[00:54:10] Kalen: Yeah. What types of notifications and things.

[00:54:14] David: Just like, how should you do a notification in the terminal and what's the, and they're also like, they start doing things like creating interesting, like stuff falling down the screen within the terminal. It just seems like they're like getting really at making the terminal experience better just because of the way that these, it's so crazy.

[00:54:35] LCI things are happening.

[00:54:37] Kalen: I never would've expected terminals to be like, I know all of a sudden this primary thing again. I mean they, obviously you use 'em all the time, but it's not like, it's not like an exciting experience to use a terminal. Like in the beginning, a terminal was an exciting experience.

[00:54:54] It was like this is all new, like this whole thing is magic. Like this whole computer is magic and all. It's like green and black glory. Mm-hmm. , All of a sudden, like, they're like the, in the interesting, like most interesting canvas for the, for computers. It's like so

[00:55:13] David: weird. Oh yeah. I'm superpowered at the terminal and they have I bought coffee through the terminal with their thing.

[00:55:20] Um, oh, that was really interesting.

[00:55:22] Kalen: Yeah, I've seen that. I've seen that. I really wanted to try it out. It looks, it's like so perfectly nerdy.

[00:55:29] David: Yeah, the coffee

[00:55:30] Kalen: was really good too. Nice. To be honest, dude, we literally need to get them to be a sponsor and like, we can do the sponsorship for free. I don't even care.

[00:55:41] But that's like, that's such dude, I'll do it for five bucks. Like, just gimme something so I can say you're a sponsor. That's like such a nerdy, like perfect Shopify fit. , It is.

[00:55:58] David: I'm, I, I wonder if people at Shopify know about it and have gotten coffee through there. Yeah,

[00:56:03] Kalen: yeah, probably. But yeah, dude, I gotta figure out how to parallelize my, Claude my clot because, and 'cause even when, a lot of times what I'll do is I'll go back and forth between two different tasks for the same project and I'll be, you know, I'll have like a little more complicated task and Claude where I'm gonna need to read the code more and understand the context more.

[00:56:23] And then I go, oh, and then I, you know, here's the small thing I wanna do that's touches a different file. Let me fire up a Claude terminal session and do it there. Mm-hmm. And even when I do that, it's like the context switch back and forth. I feel like it really, I would've really just been more productive if I just meditated for 60 seconds in between and didn't do anything else.

[00:56:50] 'cause the context switching slows you down.

[00:56:53] David: So for you personally?

[00:56:55] Kalen: I think so. For, yeah. I think for most everybody, context switching, it's a huge mental tax. Like yeah, like it's much higher than you would think it is. And there's like numbers and charts and things I've seen about that. But I think for me, more so than average, probably I,

[00:57:12] David: I was meaning like between you or Claude.

[00:57:15] And I feel the same pain with context switching. 'cause what I'm doing is like still thinking about the other thing. And I can't fully, like some days I'll have so many meetings about so many different things, and the hardest part about that day is just trying to be zoned in on the thing that I'm looking at right now.

[00:57:34] Mm-hmm. Because there's so many other things that are like floating around in my head. So yeah. The like mm-hmm. It, it feels like Claude would just have to, I feel like if my job was something like I have these tent Jira tickets I need to work on. And the Jira tickets were well-defined. Yeah. I could totally spin up like three different quads and here's the ticket details for this thing.

[00:57:55] Help me get started. Yeah.

[00:57:56] Kalen: Yeah.

[00:57:56] David: That would be like, that would be fine. But yeah, when you need to think deeply about stuff and it's like something that's brand new or not well defined, you have to spend a lot of time with Claude to work through what it is that you actually want.

[00:58:09] Kalen: Yeah. I mean, if you're not doing something trivial and like to the extent you're doing something trivial, you could fire up a thousand relatively trivial GitHub tickets in a day.

[00:58:20] The question is like, what are you actually gonna be thinking about? Like, what problem were you gonna, because if we're not solving problems, like we don't know what to do with ourselves, like we have to be actively like solving problems, you know?

[00:58:35] David: Mm-hmm. And it's a tough, so for some devs, it's like a very, very easy, I think to just spin up a few quads and they're working on completely different things.

[00:58:44] Other developers, the problem is not completely well-defined.

[00:58:48] Kalen: I must be really slow because I go really slow. If I have two things, even two things open, let alone three. And I really have been wondering if like context switching is gonna become a more value. Like a more value. Oh, definitely. I think it's really valuable to not context switch, because you can do deep work and generate a lot of value.

[00:59:09] Mm-hmm. But I wonder if AI's gonna change that, if you should actively work on like exercising your ability to switch context or something.

[00:59:21] David: I feel like that's the case. And I, I feel behind on context switching too. Like at some point they're gonna be valuable enough that you're just gonna be managing agents, I guess.

[00:59:31] And so you're gonna have to be able to provide good feedback to all those different, like, it's just like if you're a manager of like 10 people or something. Yeah.

[00:59:38] Kalen: Yeah. 'cause I, yeah. I get, it's like, personally, I think most of the time, like 80% of the time, the context switching people tend to do a lot of check slack, check email, do some code, do, do is like high, like unproductive.

[00:59:54] But it's just this like human tendency. So I've always felt very strongly about that. And then all of a sudden I'm like, wait, what if being able to switch back and forth is gonna be more valuable? And it's interesting because it's kind of a little microcosm of like, what do we do when there's no work to do?

[01:00:12] Because the AI is better than us, right? It's a, it's literally that for 60 seconds. Um, because like I'm literally, I'm sitting here, I, there's really nothing I have to do. I'm getting paid for this time that I'm sitting here and there's n absolutely nothing I have to do. So then, I don't know, maybe for some people they're totally chill with that.

[01:00:38] For me, I'm like, I have to find something to do. Like Yeah.

[01:00:43] David: I can't just watch you work.

[01:00:44] Kalen: Yeah, yeah, exactly. And maybe some, and obviously a portion of the population is just gonna completely check out or whatever, but

[01:00:54] David: I don't know. Yeah. It's like, 'cause I, I feel like I got, , a bit of a dopamine hit , while I'm typing out code, just like, oh, I wrote that line.

[01:01:02] Right. Okay. I feel good about that. Right. And now. It's like a hundred x. Okay, now a hundred lines of code just got written. And I feel, no, I feel, I feel nothing. So that's,

[01:01:13] Kalen: yeah, that's funny. Like in a way I feel like I feel more dopamine and I wonder if it probably varies by people because I've always felt like I care more about like the end result, kind of getting a thing out into the world than like the craft of it.

[01:01:36] And I think maybe some people, particularly like really good developers are sort of more into the craft. And I've really been enjoying how quickly I can get things done. And also. The stuff I can do, like I'm building a Mac app. I, I mean, it's the dumb, it's literally the dumbest Mac app on the planet. It's the simplest like hello World Mac app that's ever existed.

[01:02:06] But I literally never thought I'd create even that. So just, yeah. That's awesome. Being able to just, or like the Docker shit, like Yeah. Spin enough Docker. There's no world, like I would've, because I'd be like, that's a DevOps thing. Let's like, that's not my realm. I'm not supposed to know how to do it. Yes, some developers do know how to do it because whatever, but it's not my job.

[01:02:32] I'm not gonna do it. I just, I know I would've reacted that way, but Totally.

[01:02:38] David: I feel like I'm somewhere in the middle. I have some need to like. Somewhat deeply understand what's happening, but also I like to just get stuff done. So like, yeah, it's a bit of a battle for me. If I have Claude one shot, an app script that I need for Google Sheets and it works, I'm like, okay, I guess that's cool.

[01:02:57] It works. Yeah. I'm like, I kind of wish I would just go in and read all of it like really closely and Right,

[01:03:04] Kalen: right. Yeah, I, I think, I feel like, still feel like excited unless it was trivial, but it was, if it was kind of difficult, I'll be like, oh man, that's so cool that it got it. And then I'll be like, okay, but now I need to figure out what to do next.

[01:03:18] And it's like figuring out what to do next is also a whole thing that it's, mm-hmm. It's better to just sort of like figure out what to do next, like once or twice a day, you know? Right. Like two, you have two major things or three major things, but like all of us, there's like this. Like emotional, like, okay, now I gotta switch my focus.

[01:03:42] You know,

[01:03:44] David: another paper from the inbox.

[01:03:48] Kalen: Yeah. Yeah. So you kinda like lean back in your chair and you slowly reach towards the thing to go the next one, sharpen

[01:03:59] David: the pencil.

[01:04:01] Kalen: But yeah, I really gotta figure out what to do in between. I really want to get like the ability to work on like multiple things at the same time.

[01:04:10] Me, and I wonder if I just to like train myself to like do that. Like in the same way people train themselves to play chess, multiple games of chess at the same time, which like mm-hmm. Seems completely like I was listening to a podcast with Magnus Carlson and it seems completely inhuman, like a magic trick that you can beat 30 people blindfolded.

[01:04:34] It seems completely unreasonable, but he's just like, Hey, he just has the reps. Yeah. It's just like his brain is

[01:04:41] David: wired that way.

[01:04:42] Kalen: If you have to be at a certain level, and obviously that's a lot different than having three cloud sessions going, but like you have to be like a certain level and then you, train it and that's it.

[01:04:54] So like I'm like, I could literally like work on, like I could be working on a client thing at a solid level of productivity and I could have an entire full day's work on my own personal thing. If I could context switch like efficiently enough. It's like I could have an entire extra job working on all my app ideas.

[01:05:22] CO's like what? It's like

[01:05:23] David: legitimate, like if you hire someone though, right? Like I, I hired this person and now I need to make sure that they have work to work on and Right. I need to answer their questions when they have them.

[01:05:32] Kalen: Exactly. Exactly. Or you could spend more time on communications and be able to be twice as responsive to everyone while at the same time more productive.

[01:05:45] And that could be a cool, like that I could see the value in that. A lot of times I don't respond very fast 'cause I need to focus on what I'm doing and stay productive, but like communicating twice as fast and being twice as productive, that I could see that being a, there being a lot of value to, to at least perceived value.

[01:06:06] I think most times people think they need information faster than they really do. Like they could just review everything once daily in bulk or whatever. But I think they perceive it as being pretty cool. I don't know.

[01:06:20] David: Yeah, I agree.

[01:06:21] Kalen: Any who's it? Definitely a shift.

[01:06:24] David: You better like management by the Yeah,

[01:06:28] Kalen: maybe that's it.

[01:06:29] No, I think there's gonna be less management and by the, I could talk about this forever, dude. I think there's gonna be less management. What are you talking about? Dude? There's gonna be five. You gotta manage the quads. There's gonna be one person, billion dollar companies, dude. There's gonna be, like a five person, billion dollar company is gonna become normal dude in like

[01:06:51] David: yeah.

[01:06:52] Three to five, five person company. That five person company probably has like a hundred clawed agents running, doing

[01:06:58] Kalen: stuff. Yeah.

[01:06:59] David: And they have to manage those.

[01:07:00] Kalen: Oh, oh. That kind of man. Oh, okay. I thought you meant people manage. Yeah, yeah, yeah, yeah. No, yeah, yeah. Totally. That's the thing. You're, you're managing.

[01:07:07] But I don't know, man, like, people talk about that analogy. , I feel like it's completely different because managing people is like, that's the whole thing is you're managing the person. Like when, if you're using agents, you're using them directly. Like, you know what I mean?

[01:07:26] True. Like, I don't know,

[01:07:29] David: like it's like there's still a ton of work in like outlining the work that someone should do and then also managing the person that's supposed to be working on that. You just, you remove the person part. Yeah. But there's still the, you like providing what you should be doing.

[01:07:43] Kalen: You, yeah. Yeah. But like for example, for me, I use claw code heavily, but I have to completely un, so you could say that's an employee of mine, but I have to completely understand, and I'm a hundred percent responsible for everything that it does, everything that it outputs, which is like kind of different, , and to do that in 10 different domains.

[01:08:08] I don't know how you would do that.

[01:08:10] David: Contact switching, baby contact, switching.