[00:00:00] I've been watching more videos of the philanthropic team and stuff like that, and I saw one where he said, why is it that like virtually all engineers prefer to use cloud code even though other models do better in the benchmarks? And he's like, well, we don't optimize for the benchmarks, and other people do.

[00:00:16] When you see a benchmark, like, yeah, it is impressive on a certain level, but you have to remind yourself that if they're optimizing for them, then they're, you know, they're like over training on like these one specific things versus Yeah. You know what I mean? Who benchmarks the benchmarks? Yeah. Yeah, exactly.

[00:01:20] Kalen: Dude, I'm so mad that we didn't get that little moment recorded. That would've been the fucking perfect. Do it again. Do it again.

[00:01:29] David: I'll, I'll throw it in there at some point. Makes sense. Throw it right now. Let's see if we can recreate it. God,

[00:01:36] Kalen: that was perfect. Rock. Is this true? Yeah. See, it's, it's gone

[00:01:41] David: , Once the moment it's gone.

[00:01:42] Kalen: Once the moment's gone, it's gone. Like that's why,

[00:01:46] David: gotta hold my tongue until the record button is pushed.

[00:01:48] Kalen: Yeah. Just, just, just, you know, but the grok, the grok is this true thing is so hilarious because like I ask it that so often now, like it,

[00:02:02] David: I do you do it in the public replies or just like I Not that much.

[00:02:05] Kalen: That's the thing. A lot of them like. I've done a couple of those, but, but most of them, you know, I'm kind of embarrassed that I don't know, it's either like an inside joke or it's like Yeah. Some technology that I just, you know, I just like don't know what it is and, so I find myself

[00:02:25] David: Yeah.

[00:02:26] Yeah. I remember like, yeah, as soon as people started doing that, everyone's like, oh, this is so annoying. Everyone's got these gr is this true replies, but it's actually kind of helpful sometimes

[00:02:34] Kalen: it, it, And I'll see those in threads and I'll go, this is actually very helpful. Even, even when it seems like the person doing it is kind of being a douche.

[00:02:44] Like you get that sense. It's still like rock's reply is still like, like that's genuinely helpful. Mm-hmm. It is like, robots are better. They're not only are they gonna be better than us, like they're already better than us, like.

[00:03:00] David: It's funny, at least at reading tweets and understanding context of tweets.

[00:03:04] What's that? Oh yeah, yeah. At least at like reading all of the relevant tweets and

[00:03:09] Kalen: sense of what's happening. Well, it's funny because we've switched from using Slack to using linear for one of my projects, and, , I really like Linear. But I find myself getting like super passive aggressive in, it's a weird dynamic because on the one hand it's like, it's a great product.

[00:03:28] It's, I'm really glad we, 'cause I recommended it and then they, at first they were like, oh, we gotta use Jira. And then they were like, okay, we can do it. And I was like, okay, cool. I was happy about it. But it lends itself towards like everything's very detail oriented. So like if a ticket is real long and then it seems like there's some like fluff in the ticket or like some inaccuracy, I'm just like, I don't even want to read this ticket.

[00:03:52] I don't know. It's a weird rock. Yeah. And then so, but then the funny thing is I also been using linear MCP and with linear MCP, you can just tell it to start working on a ticket. It pulls it in and then it just as if you prompted it yourself, but you're skipping that one step of prompting it yourself and it just gets to work on its own.

[00:04:13] Right. So you can just ask it for a plan. For example, pulls the ticket and starts working on it. And it's so funny 'cause I'll look at like the way that, it's writing and talking, right? And I'll contrast it with like. The way that I reply and stuff like that. And I'm just like, wow, it's just so helpful.

[00:04:32] It's like it doesn't complain about anything at all. It just gets right into the details and stuff. I'm like, it's better than me.

[00:04:43] David: Wait, so you're using linear as a replacement for Slack as well as Jira?

[00:04:48] Kalen: like, not literally, but in the sense that we were just, , using Slack for everything. Like it was like, Hey, work on this. Okay, cool. Done. Okay. You know what I mean? But Jira is the official thing for the company.

[00:05:04] , But like, we both hate Jira, so we kind of just are like, let's just use Slack, , instead of having it goof around in Jira. , But then, well, what if we had a little more structure? What if we could use linear? 'cause I've always heard such good things about it. And so then he's like, cool. So we start using linear,, it's just like our little side.

[00:05:23] David: Interesting

[00:05:24] Kalen: thing. Yeah. It's like not officially endorsed by the company. Like we haven't exactly told everyone like that we're not using Jira. We're just like, yeah, just a little side quest type of thing.

[00:05:37] David: Well, it looks pretty cheap, like

[00:05:39] Kalen: Yeah, it's, might as well give it a shot. I think we're still on the free tier.

[00:05:41] I mean, it is an incredibly beautiful product and like, just very well designed. ,

[00:05:47] David: I was really interested in like you replacing Slack with this because if there was like, , a project management tool where your conversations also happened and then it just like.

[00:05:59] Extracted things from them that are good for context for tickets, that would be pretty cool.

[00:06:05] Kalen: Yeah. And the crazy thing is that it, we do like, like 95% of our slack, comms have moved in there, , into linear. And it's also very real time. Like, it's very obviously async, but it's very real time because , it'll show a comment immediately on the screen if somebody else writes it.

[00:06:26] If you're in your inbox view and you're, clicking through things, like you reply to three things by the time you're on the fourth thing, you see a reply come in to the first thing and you see it refresh right there. So it's like nice. It almost, it, it's like, it feels almost as real time as slack in some ways.

[00:06:47] But obviously you're just, , talking. Pro, , issue

[00:06:50] David: specific. Yeah. But all of the conversations are like on a ticket or something. So Yeah, it's organized.

[00:06:55] Kalen: And then, and linear has a concept of threads within a ticket, which sounds kind of annoying. Mm-hmm. Or like, weird, but it's like really elegant because it's like, it's, you comment on a ticket, , it just automatically, it starts a thread as a comment, but then people can reply to that thread and then you can resolve the thread.

[00:07:19] So instead of just this infinite tree of comments and sub comments, anytime you write a comment, you're picking between, is this a new thread or are we on this existing thread? And then once it's resolved, it collapses it down in the view. So that it, right. 'cause it's like, okay, I don't want that cluttering up the ticket if that little thread.

[00:07:41] Yeah. So it's

[00:07:42] David: just, it has a lot of good features, dude. It's like, that's interesting. I have to look at it. We use Jira and Asana and mm-hmm. Like they both have their strengths and weaknesses. I think one of Asana's weaknesses is that everything is like, kind of at the same level with the comments.

[00:07:59] And also I'm just always scared that I'm accidentally gonna mark something as done and then lose it forever. Right? And then Jira is just like too much babysitting, right. , Trying to structure everything to right to, and then it gets lost. Like, I'll make an epic for something and put the epic in a sprint, but you don't see epics in sprints, at least in our case.

[00:08:21] So I either need to like, dig down the rabbit hole to figure out why, or just make. Tasks on top of the epic. Right. And it's just like, I don't wanna think about it. Yeah.

[00:08:31] Kalen: I mean, and like, if I were you, I'd look into trying to use, an MCP as well, if you're doing, I mean, yeah, because like 'cause I was thinking, I was like, well, maybe I could use Jira for the stuff I need to use it for, like with NMCP.

[00:08:45] 'cause it's like, you basically are just chatting and it's just like figuring it out.

[00:08:51] David: Mm-hmm.

[00:08:51] Kalen: And it's like, I went from like, oh, this is such a cool web app, like linear and a cool desktop app. I'm like, oh, this is so nice to them being like, wait, I, and then this one guy on Twitter replied to me and he was like, I just used the linear MCP.

[00:09:07] He's like. Literally the only time I even open linear is if I'm in a meeting and I need to screen share something. And then I was like, damn, that's good. So now I'm trying to not use linear at all. I'm trying to do everything

[00:09:21] David: outta the, just straight through Claude

[00:09:23] Kalen: out of the command line. But the threads make it a little tricky.

[00:09:25] 'cause you know, you could tell it to work on an issue, you could tell it to post a comment, change a status, change an assignee. It does all that really well. But then replying to a specific thread, I think it will be complicated.

[00:09:39] David: Not out of the gooeys yet.

[00:09:41] Kalen: Yeah. Not just yet, but dude, I am all about the terminal these days.

[00:09:47] Dude. I mean,

[00:09:49] David: it's, we were talking about that last week. The EY is becoming a thing,

[00:09:53] Kalen: right? Right. Like, but I, it's like nonstop and I'm finding more reasons to do like my little email app. I'm gonna port that over to the terminal. I'm gonna like, I've spun up, like I built a, a linear CLI like for, to do things like that.

[00:10:11] The MCP was having, like all I want to do is turn everything I use into a terminal app.

[00:10:18] David: Yeah.

[00:10:18] I wanted to make, I've wanted to ask how, like when you're in it and you're working on stuff and you're like, oh man, I have these five things to do. At what point do you realize I actually should just work on a command line tool right now to make this easier?

[00:10:34] Or is that just like always there? That's

[00:10:36] Kalen: like always, that's like, that's like all the time because, and then the more that you're using command line, the more you're in it, the more like you can, so like for example, like I'm starting to have like a common format for each project I'm working on.

[00:10:53] Like I have a terminal open, I have like three tabs in the terminal. Like one is a built command, one's a clouded session, and the other one's like a regular shell for doing stuff. And then like I can flip between them and kind of know very quickly where I'm at, how to, like, basically I'm trying to figure out a multitask better because.

[00:11:14] My entire life I've tried to avoid multitasking. 'cause I thought it's the least productive thing humanly possible. Versus actually like deep focused work. And now I'm like I think it's changed. I think it's just fundamentally changed. And so now I'm like, well what am I gonna do in the 2, 3, 4 minutes waiting for the prompt?

[00:11:34] I gotta go work on something else. So you know that, so then like I'm just, I have this whole turmoil, whereas if I open Chrome, right? And I'm starting to use Chrome tabs, but it's like, not it, it's not perfect, but like,, it's not a cleaner context separation. Mm-hmm. What would really be nice if I could use virtual desktops and literally just slide between virtual desktops, but, and then have all your context in one, like your relevant Chrome tabs, your blah, blah, blah.

[00:12:05] But like, I think I. More memory. 'cause I think you'd have, I think I tested it, but it like didn't do it right or so I, like, I needed separate chrome,

[00:12:13] David: like using Mac OS spaces or whatever.

[00:12:15] Kalen: Yeah. I needed chrome profiles or something. I don't know. It got complicated. But that's like, all I'm thinking about now is how can I paralyze stuff better?

[00:12:25] Heck yeah.

[00:12:27] David: I feel like I'm so behind on that

[00:12:28] Kalen: dude. I feel like I'm, I'm so behind. Like I hear someone, talking about doing six at a time and I'm like, well, are they just vibe coding? Like zero thought whatsoever? Maybe. , And I've seen a couple screencasts and stuff, but I know there's people that are.

[00:12:44] Very effectively, like parallelizing, like people that are using get work trees and shit like that. I tried to do a get work tree and I was just I'm having trouble spinning up unique environments per, I did a work tree. I went into the folder. I get sort of how it works.

[00:13:03] It's pretty sweet how it works, but like, I don't have a way to spin up a unique environment tied to that, with Gotcha. Like a separate database. Like we just have dev environment staging and prod. We don't have dynamic environment like,

[00:13:17] David: What is a get work tree?

[00:13:18] Kalen: So a get work tree is basically what you do is you're in your repository, okay? you're In your, reviews App repository, and then you go get work tree add, and you call it, let's say, my app named Dash new feature. It creates a separate, directory that's like a sibling to your main one.

[00:13:42] So it's not inside of your main project. , It's a sibling to it, and it, it automatically sets it up on a different branch and then, but it's not a separate clone. So like from within the work tree folder, you see everything that's going on in the other main folder. So like, if a push or a commit happens, they're like looking at the same history or this, like the same git , history or whatever.

[00:14:12] , But it just sort of magically like works. Like that's what it's basically for.

[00:14:17] David: Cause like the, the traditional approach to that is just having different branches, right? But work trees maybe are better for if you have separate branches, but you're one person or something.

[00:14:27] Well, so

[00:14:28] Kalen: it's separate branches and you can have both checked out at the same time.

[00:14:33] David: That's blowing my mind up. I'm gonna have to look into that

[00:14:35] Kalen: because you're working on one feature branch here, and normally you just switch branches whenever you wanna work on a new feature. But, but then mm-hmm. If Claude is running against that branch and then you Gotcha.

[00:14:48] You wanna go work on another thing, you need to be, you need to go into the other branch and in another place. So the, the simple way of doing this is you just clone the repository into another folder, check out a different branch. Mm-hmm. But if you do that. They're separate. Like things get, doesn't

[00:15:07] David: know about it.

[00:15:08] Yeah.

[00:15:08] Kalen: So, yeah, exactly. So

[00:15:11] David: interesting.

[00:15:12] Kalen: Yeah, dude. So man, and I've had some moments dude, where I'm like cruising through multiple things at once, like efficiently without that whole like, switch context, scratch your head for four minutes trying to remember what you were doing type of thing.

[00:15:31] Mm-hmm. Like I've had some of those flows where I'm like, dude, this is awesome. And then sometimes it's totally the opposite, you know, sometimes it's like, oh my God, this is, I should just focus on one thing. 'cause I feel like I'm just wasting time context switching. But it's like I'm trying to, , improve that whole thing.

[00:15:52] David: That muscle. Yeah. Yeah. Do you listen to like crazy cyberpunk techno music? Oh, yeah. When you're managing all these AI robots? Oh yeah,

[00:15:59] Kalen: yeah, yeah. Dude, that's, you gotta have EDM going at all times.

[00:16:04] David: Step one. What? Get the bass flowing. Well, yeah. What do you listen to? I jump around. Today I've been listening to K-Pop Demon Hunters because

[00:16:15] Kalen: my kids love that show.

[00:16:18] David: Did you see they're playing them in theaters? No, they, they started, , like getting theaters and they're everywhere. Like out here in Missouri. They're, that's hilarious. They're playing it everywhere. I'm going with my daughter tomorrow to see it. It's like a sing along, so That's wild. But yeah, I, I usually like, when I need to get into focus mode, it's either heavy bass music or some sort of like progressive metal, like crazy rifts going right.

[00:16:44] Kalen: I didn't know you looked 'cause I haven't really listened to metal, but when you said that and remembering your whole vibe, I'm like, oh yeah, a hundred percent. Oh yeah, that makes sense. That sucks. That's cool. Yeah, like, dude, how about people that don't listen to anything while they work

[00:17:04] David: is, oh, they must be serial killers.

[00:17:06] Right, right. Like be at Duke, like people who listen to nothing while driving. Oh, how do you do that?

[00:17:14] Kalen: I play music where I play pickleball and like, like one time I was there, somebody had a speaker, I was like, ah, that sounds good. So then I eventually got a speaker, DJ

[00:17:22] David: Kayla, I

[00:17:23] Kalen: got, yeah.

[00:17:24] Every once in a while somebody like, complains, it's too loud or something, and it's like, or, and, I brought it for a while and then I stopped bringing it because one person complained and then, yeah. And, and then feel bad. When I stopped bringing it, people kept asking, where's the music?

[00:17:38] Where's blah, blah, blah. Anyway, so then I started doing it again. And it's like, p and especially these fucking tennis guys that are playing pickleball now, and they, they're fucking fancy and like, they're like, yeah, this is a supposed to be a gentleman's sport, you know, there's no music. I'm like, go fuck yourself too.

[00:17:58] David: Pickleball is a gentleman's sport. Yeah. I feel like pickleball, I feel like the whole thing around pickleball is it's. Casual Exactly. But maybe, maybe not.

[00:18:05] Kalen: Well, but No, but his thing was that tennis is a gentleman's sport and, and he's basically treating like, I feel like they have the right to impose their tennis culture because, you know, it's tennis is sort of like better than 'cause it, it's adjacent pickle ball.

[00:18:20] David: Yeah. It's,

[00:18:20] Kalen: it's adjacent, but it's also better like there, you know what I mean? Like, like nobody respects pickleball, , 10 tennis players get a little bit of respect. So like

[00:18:29] David: I understand that's what makes pickleball punk.

[00:18:31] Kalen: Yeah. Yeah. Puts punk rock, dude. It's fucking That's right. But it's also too popular.

[00:18:35] Don't listen to

[00:18:35] David: anybody telling you to turn the music down. Yeah, dude.

[00:18:38] Kalen: It's, it's, it's bomb for the vibes. So what else do we got? What? Oh, toss posted ACL i. He created ACL i, speaking of CI called. Try that. Lets you organize your I saw that. Yeah. Yeah. I was like, this is freaking awesome. And then when I saw that, yeah, that's interesting.

[00:18:57] I was like, shit, I do have like a million folders in my main repos directory and it's a hot mess. And like, I'm like, dude, I need this. And, uh. So that was smart.

[00:19:10] David: Yeah, I was, I went in and looked at the GitHub repository a little bit 'cause I like, I wasn't sure I understood. And there's like, so the main thing is that it creates new directories, but there's like all these really nice things around it, auto dating and like auto switching between them and stuff like that.

[00:19:29] And, and it also has its own little tuy. So it's, it was pretty cool. That's crazy. I haven't tried it out.

[00:19:34] Kalen: Yeah, I took a look at it and just literally the search functionality was, I was like, Ugh, I get it. Because like, it, it, instead of a regular, files, , directory search, you're tabbing into it, you're tab auto completing, so it just, you have to type the first part correctly, you know what I mean?

[00:19:52] Mm-hmm. Whereas it just, you know, the search just searches anywhere in the, in the directory name. So it's like, ugh. It makes it so much easier to like find stuff. But I haven't fully integrated it into my workflow. .

[00:20:04] David: Yeah. I don't find myself making new directories that often. I'm mostly like in directories that I already made.

[00:20:10] Yeah. Or I'm like, okay, I need to do this new thing and I just make a new directory and call it something. Yeah. , But if you're like doing research stuff a lot Yeah, definitely. Seems like it makes sense.

[00:20:20] Kalen: Yeah. It's like it always feels uneasy creating a new directory. 'cause you're like, wait, I'm kind of committing to this thing.

[00:20:26] Like do I really need to create this or can I, like, should I be doing something else right now?

[00:20:35] David: now you got directories per date, all the directories.

[00:20:38] Kalen: And then I was like, wait, that sounds horrible. I was like, by date, now I have to remember the date to go into it. And it's like, no, you search on it, it shows you the date.

[00:20:46] And all this stuff, dude. So I gotta This guy Toby, I think he, I think, I think he has some talent. I think he's going places.

[00:20:53] David: I think he go places.

[00:20:54] Kalen: I mean, call me crazy. ,

[00:20:57] David: I've also been seeing a lot of the like, Shopify, like the non dev teams posting their vibe coded work. It's making me, it's pretty cool.

[00:21:05] Me

[00:21:05] Kalen: so

[00:21:06] David: angry

[00:21:06] Kalen: how good

[00:21:07] David: they're,

[00:21:09] Kalen: it's like, it's absolutely killing me. it's So cool, man. Like I saw,

[00:21:15] David: it makes me excited it feels like there's just a lot more software coming out lately. And like, it could be that a lot of it is bad, but that might mean that overall there's still more good software coming out.

[00:21:30] Like if there used to be 20 projects coming out a week and now there's a hundred. Right. But 50% of them , are horrible. That's still. Like 30 more. Right. So like 150% more good projects.

[00:21:42] Kalen: And even aside from the net new ones, the existing ones are getting better too. Mm-hmm. Like all my software that I'm writing is better quality now, like significantly.

[00:21:55] And you know what I mean? And like, I'm sure everyone, I'm sure most people's, , improvements are even better than mine and, you know what I mean? And like, yeah. You got talking Stewart with Revo out here destroying the game, dude. Yeah, he's at the top. Did you follow him yet? I think we talked about.

[00:22:12] I think so. Okay. We talked about you, I talked to you about this a lot. You're like, who's Stewart? I'm like, dude. Yeah, he's, he had another hot tip dude, which was an auto Cloud MD files, basically an auto, an agent that automatically updates your Cloud MD files as it's doing work, it automatically detects if there's something that should be reflected in the new, in the Cloud MD files based on what it's working on and makes those changes, which I'm just, because I'm spending a lot of time trying to optimize my Cloud MD files and it's like,

[00:22:48] David: yeah.

[00:22:48] Kalen: It's like,

[00:22:49] David: I need that.

[00:22:50] Kalen: Yeah, it's like, and I did some separate files per directory. I, you know, I feel like it's helping. And then I have sort of a workflow now for if I see it do something I didn't think it should have done instead of just like clicking. Yes, yes, yes, yes. I, let it run.

[00:23:07] I go a separate tab, I open a cla and I talk to Claude about Claude. You know, I go, I go, Hey, this other Claude did this, but I didn't think it should have because I thought that was already in the cloud MD file. Can you look into it? And then it goes, yo, yeah, it's because you need to add this section. And I'm like, perfect.

[00:23:25] Cool.

[00:23:26] David: You know what I mean? That's cool because then it doesn't taint your context with the, uh, or your conversation with other one. Exactly.

[00:23:31] Kalen: Otherwise it throws nice.

[00:23:33] David: Yeah.

[00:23:33] Kalen: So that's like a little, that feels

[00:23:34] David: like something that should be in Claude code, like the, not the, not the separate Claude, but the be better at updating Claude md Please.

[00:23:42] Because like your context window's gonna run out eventually. That's, yeah. Like your life

[00:23:46] Kalen: depends on it. Like it should be doing that all on its own, I mean. Mm-hmm. And there is this memorized command if you hit the pound key. And I kept seeing it in the like, tip of the day, like hit the memorize. And I was like, what does that mean?

[00:24:00] And then I finally used it today, and all it does is takes whatever you type and puts it at the bottom of your Clot MD file. So I don't understand why people would use that because you you need some organization to it, so like, probably whatever you said is related to some other set of bullet points somewhere else.

[00:24:19] So you want it put in the correct context, you know what I mean? Right. I thought it was gonna automagically do that,

[00:24:25] David: I feel that way with a lot of those like question mark or whatever commands that you can do in Claude code where it's like you can list out something or you can reference specific files and I, I feel like maybe my expectations are high on what it should understand and just be like, I don't want to tag a particular file, I just wanna talk to you about concepts and then like, you know where that is.

[00:24:47] Kalen: Right, I find that I'm not having to tag files too much. Okay. Do you have to tag files a lot or does it

[00:24:55] David: just fine? No, I've never done it. I just saw that it's a thing that you can do. Oh yeah. I don't know why I would do that.

[00:25:00] Kalen: Right,

[00:25:00] David: right, right,

[00:25:00] Kalen: right, right. Yeah. Well it's funny 'cause using cursor, like you had to, to do that, like it didn't know any really anything about anything un unless you tag the right file.

[00:25:12] So I got used to doing that and then Gotcha. It was one of those habits where like, I felt like I had to do in Claude and then after a while I was like, oh, actually, I mean, it might speed it up a little bit, it's so much nicer to be able to think at a higher level..

[00:25:27] I used to have this habit. Anytime I got ready to work on a feature, I'd be like, okay, I'd crack open the editor and I'd be like, okay, what, what file is it in? And then your brain is kind of traversing your mental model of the project kind of slowly, you know, like cracking off.

[00:25:43] Mm-hmm. You know, the dust webs and, and cobwebs and stuff like that.

[00:25:47] David: What was happening in here.

[00:25:48] Kalen: Yeah. And then you get to, and you go, oh yeah, it's in this file. And then you open the file and then you again, that was another habit that I was, when I worked on a new feature in Clot, I was like, oh, which file is it in?

[00:25:58] But now I don't even do that. I just say, Hey, there's this thing that needs to do this, and boom, let

[00:26:03] David: it, you know, what I kind of miss is, sorry, I'm bringing up Magento, but. like Trying to figure out where we gotta do it every time where Yeah, dude, we do, trying to figure out where something actually happened and like traversing the class and like function trees and like command click.

[00:26:19] Okay, now I see what's happening here, but now it calls this function command click. Yep. That was kind of fun. Like a little bit of a little scavenger hunt. Yep. But I, like, I'm scared of how Claude would've tried to like, that would be so many files to go grip and read the first 100 lines up.

[00:26:39] And you wouldn't get anywhere with that.

[00:26:40] Kalen: Dude, that's a very good question. I talked to my friend Jerry, he's doing Magento and stuff. I mean, I saw some tweets by, some Magento people. I think they're, I think they're hit, hitting Claude hard, are they, I think just doing it. Think so.

[00:26:54] I don't know. But that is a good point you mentioned. 'cause I mean, that's a lot of. Files to

[00:26:59] David: read. Yeah. Big files. Like big classes.

[00:27:02] Kalen: Yeah. No, I, now I'm extremely curious about that, but, we'll have to get a Magento guest on the show for, uh, for old time's sake. We shall. , Yeah, no, it's crazy. , Crazy times.

[00:27:14] Crazy times

[00:27:17] David: crazy. Hey, I think,, not on the list, but, . I am pretty sure since last time I joined the shop dev Alliance, I'm in there now. Oh,

[00:27:26] Kalen: that's right. Hold on, hold on, hold on.

[00:27:31] David: Thank you. I don't know if you could hear

[00:27:32] Kalen: the clap coming through.

[00:27:34] David: I owe it all to you. Kay. Round of

[00:27:35] Kalen: applause for Mr. Robinson. Everybody. , Hey, don't mention it. You know, it's, you know, it's, it's what I do. You know, I mean,

[00:27:41] David: I

[00:27:41] Kalen: may

[00:27:42] David: call someday I'll get you into some very exclusive community.

[00:27:45] Kalen: I may call you and ask for a favor one day, and I'll expect you obviously to, uh, return the favor, you know, no big deal.

[00:27:53] David: Any, anyway, I can, if you want to like record a podcast or something, we could do that.

[00:28:01] Kalen: Nah, that's cool, man. How are you liking it? Are you in there every day? Are you like all into it?

[00:28:05] David: Yeah, I've been over and checking on it and reading some stuff. There's like really cool things that people are posting in there.

[00:28:10] Nice, nice little, nice little tables of like, here's how the, , dot length works on all these different objects and it's just like, ah, I never did that, but interesting it out.

[00:28:20] Kalen: Nice. That's awesome. Yeah I haven't been in there like really in there, like looking at everything. Um, I usually pop in for quick message here and there, but we gotta start a little sub-channel or something in there and just like, I don't know, just have like a couple bros to just shit post.

[00:28:38] Yeah. On a constant stream. Ba I need that in my life. I need like a good, you know, GIF only. Yeah. Dude, speaking of slack, it's so brutal 'cause I'm doing a little, bit of work for Shopify Flow but, and I can't talk about any of it, , which it sounds like I'm doing some kind of a humble brag, which I slightly am, but.

[00:29:03] It's like, it literally drives me crazy. 'cause you know how I am. If there's something I think is cool that I'm working on, it's like I want to post about it. You know? It's like a, yeah, it's like an automatic instinct for me. And so I was doing some stuff. That I thought was cool and I was just like, oh my God, I wanna, so I literally, I messaged you and actually, I did not message you, about it, but

[00:29:31] David: I was like, wait, so you're talking about this thing that you can't talk about? Got it. Which I would, which

[00:29:36] Kalen: I definitely did not message you. This is a public show. We do have a large listener base. but here's the thing I wanted to say is that I have the worst of all worlds.

[00:29:48] 'cause what I realized after seeing all these cool ass vibe code things that they're sharing internally and every once in a while , they post one and like Gil had tweeted about the fact that they're doing so much cool internal stuff. What I realized is they don't need to share in the outside world.

[00:30:04] 'cause they have this gigantic 6,000 person universe. That they're sharing everything they're doing with, you know what I mean? Because like, if you work on something and you can't share it until Yeah, it's public like nine months later and then blah, blah.

[00:30:20] You know what I mean? Like Yeah. The feedback loop is energizing, and I'm guessing they have a cool, like internal, , feedback loop.

[00:30:29] David: I feel like everyone, , it just gets to talk about everything internally at Shopify. Yeah. And so when there's like a little bit of a, a leak, it's hard to contain it.

[00:30:39] Yeah. Right. Because you're like, oh man, this is cool.

[00:30:42] Kalen: Right, right, right.. it makes me so angry that I can't just work there because, I'm chasing this stupid ass, dream of, getting that sweet internet money and, even though I, I have a realistic shot at it, most likely.

[00:30:58] But

[00:30:59] David: there are many ways to get internet money, and I feel like, \ the thing that I've been hearing lately is you just gotta keep trying. And I, I feel like you're good at that.

[00:31:06] Kalen: I am very good. Yeah. I mean, I, yeah, so, at least at trying, but I feel like I would just be happier as like a human being and I'm sure make the, you know, more than enough, I mean, you probably have to work there for 20 years to have enough to retire, but I mean,, I feel like it would just be such a cool ass place to work.

[00:31:25] David: I agree. Their offices are beautiful too.

[00:31:27] Kalen: Oh God. Their office. Like, it's literally like I saw harsh tweeting about it and saying that, I think he said he like moved somewhere just so he could go into the offices and I'm like, yeah, that is something. If I was single I would do that.

[00:31:40] Literally like Right. Like that's

[00:31:43] David: how

[00:31:43] Kalen: fun would it not happening now in my

[00:31:45] David: life.

[00:31:45] Kalen: But that's how fun it would be to just be able to pop into a Shopify office. Like whenever you want. Like never have to, but you can only, you only go there if you want to.

[00:31:57] David: Damn. Yeah. That would be so fucking cool. Like to go in and show people something cool that you've been working on in your Yeah, dude, like Dungeon or whatever.

[00:32:05] Show up, everybody's chill.

[00:32:06] Kalen: Hey guys, want to see this thing I built? Yeah.

[00:32:09] David: Be like TV show, dude, one of the things I miss from being in the office is we used to do these like Hack Fridays where we'd just work on something that's fun that we haven't really set aside time for. What company was that at? , We did it at aoe.

[00:32:24] Okay. That's, uh, because I was gonna say every once in a while

[00:32:26] Kalen: that sounded familiar. I remember, I think I remember, yeah. You guys talking about that.

[00:32:31] David: Yeah. But the, like, the energy isn't there. I think if you're all kind of like in separate spaces, I mean, you could still make it happen. It's just

[00:32:38] Kalen: different.

[00:32:39] I mean, but I think you can make it happen. I was actually talking to somebody about, I don't, I don't think you ever saw that little Magento community I started. It was just kind of random. It was a, it was, uh, it was like the last thing I did

[00:32:52] David: That was a Slack thing too, right?

[00:32:54] Kalen: Yeah. Yeah. And Yeah, it was pretty cool. I ended up handing it off to PJ to run 'cause he wanted to run it and so I was able to, to. You know, I was making a little bit of money. I was able to just hand it over to him. Didn't try to sell it or anything like that. Nice. But we used to, we would do like a Friday, like I, I wanted to do something kind of chill, so I'd be like, Hey, let's have a beer on Fridays and like, talk about, you know, instead of the typical like, somebody's gonna do a talk and we're gonna listen to the talk.

[00:33:20] It was like, we'll do a hangout and then we'll have a beer and like somebody, like, we'll have a topic and like somebody will be like, not really giving a talk, but just kind of like talking about it and just have a conversation with everybody. And yeah, it was kind of fun. It was kind of fun. I kind of want to maybe try to do something like that, but just for fun.

[00:33:41] That's cool. Not trying to monetize.

[00:33:42] David: Yeah, like I could do like a virtual version of it too, like a little mini hangout. We try and do that sometimes at Ana, we have like a. Virtual scavenger hunt thing, which, you know, like, maybe that wouldn't be what you would do, but just like,, virtual happy hour and talk about stuff that happened.

[00:34:01] Yeah. Kind of like what we have here, but if you want, you can like, have your camera on and just meet a bunch of people.

[00:34:06] Kalen: Yeah. Yeah. That would be kind of cool. I really like, for podcasts, I wish I was better with like, multi person. 'Cause it would be really fun to grab people, and pull 'em in.

[00:34:18] But I just, every time I like the, just the managing the lag on an internet call and like, without losing, like how many times have you talked to somebody and you're just like, oh, oh, what'd you say? Like literally it's just like so much of that, that it sort of zaps your energy, whereas like.

[00:34:36] When you're used to talking together, you just, versus just one, like one person and one other person, you know what I mean? It's like, I feel like it, the ability to have a good conversation, like goes down exponentially just because of all the waiting for someone to talk gaps.

[00:34:53] David: Gotcha. Yeah. I tried to listen to some,, Twitter, I forget what they're called, like spaces, is that what it is?

[00:35:02] Mm-hmm. And yeah, it feels like that where there's like a bunch of dead air trying to wait for someone to start talking. Yeah. There's a stage so that it doesn't get too rambunctious, but then you have to wait for that person to get to the stage. Yeah. Like unmute themselves.

[00:35:15] Kalen: Yeah. It's just, it's just horrible.

[00:35:17] I can't wait for technology to kind of fix that to where you could have, like, it's something about literally just the delay of however milliseconds it is. Mm-hmm. That on top of the fact that it messes up the interpersonal dynamics of like, should I keep talking because I'm important enough to keep talking or should I wait and wait for that?

[00:35:39] , It's a weird, it becomes a weird, power dynamic. And then I'm like,

[00:35:43] David: too patient. Yeah. Like, should I, like, if I'm just too nice about it, then I'll never say anything, I

[00:35:48] Kalen: guess. Yeah. And then it, it somehow throws off the chemistry. Like, but I don't know. , Maybe I'm just being stupid about it. Dude, notebook, LM is fucking, do you use a notebook?

[00:36:02] Lm

[00:36:03] David: I haven't like, found a reason to use it. I've seen some of the stuff that you posted about it, like the, the topics that we covered, but Yeah.

[00:36:11] Kalen: Dude, I talked to my friend Eric about it. He's taking a master's degree, just for the heck of it. Like he's retired. So sweet. And he's telling me all the stuff he's doing with it.

[00:36:22] Pulling all this information in. You can ask it questions, you can summarize it. Like I think some people are using it for Scope doc, like basically that thing that we talked about of like, imagine you could have a scope, bot that knew your, project scope and you could just mm-hmm. Throw documents at it and just ask it a question and it answers it accurately.

[00:36:44] That's what it does. It's

[00:36:46] David: okay. So it's not just the like podcast style. Have a conversation about a bunch of stuff.

[00:36:51] Kalen: It has all these different modalities and I think they added video recently. So it can take video input, it can do video output, it can do audio, it can do text output, like it can ingest.

[00:37:03] Lots of different, types of texts and data sources, and then it just, you can do mind maps on it. Like I'm, now, I'm looking for more use cases to start using it because I, so I did the, our podcast transcript search on it. And dude, I've tried three different ways to build, into the website, like a good AI search using like a vector database using open ai.

[00:37:28] Mm-hmm. Blah, blah, blah. And it always sucks. The one question I wanted to ask is what are the Shopify apps we've talked about? Because we've talked about a bunch. It's kind of interesting to see that information in aggregate, and it's like, I remember some of them, but when I saw the final list and I looked at it, I was like, oh yeah.

[00:37:45] Oh, I had all these like, oh yeah. Type moments. Mm-hmm. And, so I put it in and it just absolutely nailed it. And I just was like. This is so crazy how good it is. And also Eric mentioned that it's different from other ais and that it's actually bounded to the information you put into it.

[00:38:08] So it, that's why it doesn't hallucinate. I mean, it must have some underlying, obviously model. It knows stuff, but like, yeah, somehow it's bounded, so like it didn't hallucinate anything. Like, not even 1%.

[00:38:22] David: Like Yeah, I didn't see anything weird there. It was crazy. I saw like a little bit of, there was, uh, something that was like slightly like it could have been because the transcript, uh, wasn't perfect or something like that.

[00:38:35] Kalen: The other funny thing was like, it mentioned Shopify flow and it itemized everything we said about Shopify flow, like, like 12 different things. And in my head I wasn't even thinking of that as an app. I was just thinking of it as like it's native. But it like pulled it up and correlated all those different things.

[00:38:56] I was like, holy crap.

[00:38:58] David: I gotta give it a shot. I feel like, , I don't trust, , maybe trust isn't the right word, but I've been trying to use Gemini and Sheets Uhhuh to help me with stuff like I, I like, rather than Google search for what the formula is that I'm looking for to do this, like pivot or whatever.

[00:39:13] Right. I just wanted to ask Gemini and I'm like, I have this data in this tab and this data and this tab. Help me, pull , the column from a over into this column over here Right. If something else matches. Right. So like standard pivot thing. Right. And then it, it says something like, I can't do this because I don't know anything about that other tab.

[00:39:35] And I was like, it's literally like, okay, maybe instead of calling it a tab, I'll call it a sheet, but it's there and then it's, and then it wrote a function Yeah. Or a formula. Yeah. And it didn't work. And I was like, this isn't working. Yeah. In order to fix this, I had to rearrange the columns. Yeah.

[00:39:54] Why didn't it work the first time? Yeah. And it like, it just doesn't, and it doesn't feel like it's super knowledgeable.

[00:40:00] Kalen: And by the way, I hate Google. Like I hate, the Google console interface. I hate the, all the models they're coming out with. It's like, why? It's like, it feels like they're just copying everybody else.

[00:40:12] I feel like they're sort of evil as a corporation. I think their models are all stupid. Like this is so much, I think better than ev every other product they have and like you and it's unique in the market that I don't think anything comes anywhere close. It's

[00:40:30] David: like, and they must be doing something special.

[00:40:32] Like what you were saying, like you were saying the bounding, like that must be

[00:40:35] Kalen: Yeah,

[00:40:36] David: it's super important. Like some sort of new thing.

[00:40:38] Kalen: It's got some magic sauce in there. Like I think Gil even tweeted, I think he said that, Notebook, LM is the best AI product on the market and.

[00:40:48] I was like, well that's a lot to say because I mean, cloud code I think is like light years better, but

[00:40:53] David: Yeah.

[00:40:54] Kalen: But it is incredible. Like I'm trying to find more use cases for it. And also 1 idea I had was, you know, I want my kids to use AI more, but, you know, because we're worried about it, you know, getting inappropriate content.

[00:41:07] Mm-hmm. Even with the age, whatever, I don't know what age, like, base filters they offer. It just feels like those, things never work. There's always like, you know, stuff you don't want them to see that they see kind of a thing. Mm. yeah, you know, I thought,

[00:41:21] David: yeah, I'm gonna have to check out Notebook lm.

[00:41:23] 'cause I, I am, I'm really interested in seeing something that's like new versus all the other, like head-to-head LLM model. Yeah. Like doing work on all of these, crazy scientific. Things, and that's how we score them. Yeah. Like that's not the way people typically use those things. Right. And like, I just want help writing a very simple formula in this Google sheet and that kid doesn't work.

[00:41:49] Kalen: I was, I've been watching more videos of the Anthropic team and stuff like that, and I saw one where he said why is it that like virtually all engineers prefer to use, cloud code even though other. Models do better in the benchmarks. And he's like, well, , we don't optimize for the benchmarks and other people do.

[00:42:05] when You see a benchmark, like, yeah, it is impressive on a certain level, but you have to remind yourself that, they're optimizing for them, then they're, you know, they're like over training on like these one specific things versus Yeah. You know what I mean?

[00:42:18] David: Who benchmarks the benchmarks?

[00:42:20] Yeah.

[00:42:21] Kalen: Yeah, exactly. But like, Elon is always posting about the benchmarks and it's like, ugh. I feel like, you know, like on the one hand I sort of get it, like he needs to like grow this thing and the rate he's grown, it is unbelievable. Like, I think even like the Nvidia guy, Jensen Wang said that like he did something that usually would take a year in 17 days.

[00:42:48] Just,

[00:42:49] David: yeah, that was the big, , data center

[00:42:52] Kalen: build. Yeah. And like to hear Jensen Wang say it is like what? Um mm-hmm. Like the Nvidia guy. So like, , but at the same time I feel like, the whole like soft porn nature of like the grok chat thing, like, I don't know if you've seen of those videos, but it gives you this like, anime girl that like

[00:43:14] David: video Oh yeah.

[00:43:15] Chats. And I saw companion mode or something that it's called. Yeah. It's,

[00:43:18] Kalen: it's like, it's definitely weird. But, you know, I was thinking to myself probably he's, he's growth hacking dude. He's growth hacking. 'cause he doesn't advertise. He does things that get attention, so I think he did that probably 'cause he thought it was interesting and it is interesting, but probably 'cause he knew that would get a lot of eyeballs and like Yeah.

[00:43:39] 'cause he really believes there's an existential risk and he's like, I need to build the good guy version of this. And it's just crazy how unstoppable he is, dude. Like, as a, an entrepreneur, like, it's fucking crazy. Yeah.

[00:43:56] David: It's a lot of stuff. It's unreal. I feel like, I still have Elon blocked on Twitter, but I, I feel like he's said like fewer,

[00:44:05] Kalen: you're one, I guess we've never got to a deep enough conversation on the, you're one

[00:44:10] David: of those.

[00:44:13] It was a while ago. I don't know if things have gotten better, but you're one of those man, he was kind of annoying me for a

[00:44:17] Kalen: while. Gotcha. Fair enough.

[00:44:20] David: Because you could always, I mean, I, I'm on Twitter every day. Yeah. And that's what I did. I muted. Oh, okay. I thought you blocked. I was like,

[00:44:25] Kalen: that's a lot,

[00:44:26] David: dude.

[00:44:26] That's Oh yeah. Because Elon's gonna see my tweets. That's,

[00:44:31] Kalen: you know, you may, you never, he replied to Toby, Toby replied, , did he to something on the new model YL and Toby said something like, it'd be good if there was Starling hotspots or something like that. And Elon replied, good idea. I was like,

[00:44:44] David: damn.

[00:44:45] Kalen: Oh, nice. That's like Toby's version of like when he replies to us. Yeah. Although not probably, I don't know. Yeah, probably. I mean, you know, that's an interesting point. Like as insanely smarter that Toby is, than like you or I. And even though they're both billionaires, I bet Elon is that much smarter than him probably.

[00:45:10] Which is like crazy

[00:45:13] David: and Elon knows it. I think that's part of the problem. There's levels most of the time

[00:45:18] Kalen: he knows it. We could do a whole spiel on that, oh, can I do a quick advertisement, if you wouldn't mind. Do it. All right.

[00:45:27] So MedSync is my little CLI tool for syncing stores from one shop to another. I built it out, and actually I created, can you

[00:45:35] David: sound a little more excited about this

[00:45:40] Kalen: MedSync? You need to sync your data. There we go. Um, So basically I did a v two of it that's like better you, it basically follows the model of the SQ lite database that like Shopify's working on.

[00:45:56] Um mm-hmm. But I promise I didn't like consciously steal the idea. I literally forgot. I literally heard about it. I forgot about it. Then I started working on it again 'cause I kept needing to sync data for test stores and so then I was starting to work on it again, and then, I asked Chad, GBT, well how should I architect this because of blah, blah, blah, blah. I need to keep track of state. And it was like, use a SQ lite database. And I was like, oh, that's a good idea.

[00:46:19] Let me do that. And like, I like four hours later or something, I was like, why does this sound so familiar? And then I was like, oh yeah. So then I googled it. I was like, where's their, I remember hearing about their CLI, I Googled it. I couldn't find it. I was like, wait a minute, did I imagine that?

[00:46:38] And then, and then I posted about it and then people reply. They're like, oh yeah, it's not out yet, but you know, blah, blah, blah. Anyways, so I redid it. Guess

[00:46:47] David: I gotta release mine.

[00:46:49] Kalen: It's bigger and better. I, spent a bunch of time on it. It does, it sinks everything. It has a whole deferred resolution of dependencies model.

[00:46:56] 'cause like the V one, one of the problems with it is I kept special casing every time there was a certain reference meta field. And, , it, the code just got weird and it was like, I got stuck on this big sink and it just, , it was like I was patching, patching, patching, and it was just stupid. So I did it with a different model for more flexibly deferring resolutions.

[00:47:16] I did like a bunch of dynamic payload size to like optimize your query costs and stuff like that. And

[00:47:22] David: anyways,

[00:47:22] Kalen: oh

[00:47:22] David: yeah, I saw you posting about that. That's cool. Yeah,

[00:47:25] Kalen: so I kept the repo private for now , and I'm just like thinking like, you know, maybe somebody, you know, might want to pay a few dollars, you know, to get access to this exciting new tool.

[00:47:38] And then I could do, so that's, that's what I'm trying to do. I'm, I'm going door to door trying to sell my, my c my CLI. Yeah. I'm a door to door traveling. CLI salesman.

[00:47:48] David: This is how you get acquired by Shopify.

[00:47:51] Kalen: Right. So anyway, kind of excited. Kind of excited.

[00:47:55] David: That's cool. So it's, uh, it's like the holy grail of actually s sinking all the data from one store to another.

[00:48:01] Exactly. Also resolving all of , the dependencies of like, references to meta objects and

[00:48:08] Kalen: all the stuff stores in a local database for fast lookups, because otherwise you're having to go back and forth to like, fetch IDs and look them up and it gets really slow. So like, having that database layer as a mapping, like keeps it all running smoothly.

[00:48:22] And then also interesting, the first version of it, it was like you had a bunch of sub commands. So like if you wanted to sync meta fields, you'd be like, metas sync meta fields, and then if you wanted to sync products, you would like do that separately. And then like my friend called me was using it, he was like, I just want to run it once, like the default mode out of the box and have it do everything in the correct order.

[00:48:43] And I'm like, well, yeah. And then like I, over the past few months, that idea has like recurred to me and I'm like, oh yeah, obviously, , you just wanna, so it, it works like that now. It just like goes nice. Goes to town. But yeah. Any who? No big deal. It's fine, dude. Okay, it was so weird. Like I had a day like, you know how when your productivity drops, , or if you feel like your productivity, like you feel like you're not getting anything done.

[00:49:13] Do you get like mildly depressed about it?

[00:49:16] David: Like not. Oh, definitely. Right. I was so, like, our, air conditioning broke this weekend and I was so hot for the first three days of the week and I was like, I could tell I was just not as effective. Right. And I hate myself for it.

[00:49:31] It's,

[00:49:31] Kalen: it's the absolute, yeah. It's, and I know exactly what you mean. Like it's crazy, the degree to which literally just the temperature destroys your abil your like, desire to even mm-hmm. Think. Mm-hmm. It's like, mm-hmm. It's insane. , So you had a day like that earlier this week. Oh, but dude, it was even worse because my productivity is getting so much higher.

[00:49:57] Like I said, when it's going well, you know, or even when it's going like medium to, well, like, I'm in a range, but like the, one day this week, dude, Claude was super. Like it instead of maybe normally you wait an extra couple minutes, it was like, I'm waiting 10 minutes and it's hanging there. And then it eventually like started working.

[00:50:21] I dunno, there were some other things, but like, I feel like because my productivity dropped by so much from like a higher high, I was like very disturbed about it for the rest of the day. Yeah. It's like making me think like with the superpowers that AI gives us, like we're gonna be so dependent on those, that losing them is gonna feel like very painful.

[00:50:46] David: And that's when the CLO subscription cost goes up to $3,000 a month. Literally. Do you, after you've had a little taste,

[00:50:54] Kalen: dude, I feel like I would be lost. Without it. Like, I have this whole new future I'm imagining for myself because I feel like completely, like I have superpowers and yeah, , it's like, I feel like it's just, it's, the future's just gonna get better and better.

[00:51:12] Like if they cut this thing off and it'd be like, oh, you just have to do regular programming. Like, it would,

[00:51:20] David: you mean I gotta type all these characters,

[00:51:23] Kalen: dude. and It's not that I'm not working, I'm actually working more, I thought I would just want to kinda sit there and just chill.

[00:51:30] Like let it prompt. But it's like, you wanna be doing something, you want to be challenging. Mm-hmm. Yourself. So then you find more and more things to like challenge yourself with, with,

[00:51:41] David: have you thought about, running some sort of local model just to make sure that you're never without ai.

[00:51:46] Kalen: Okay, so the thing I don't understand, like I looked at that, but like, I think because I only have eight megs of memory and I'm about to get a new computer, but, , I don't even think you can run like, , open AI's local model with, I think you need 16 or 32 or something like that.

[00:52:02] I'm like,

[00:52:03] David: yeah, you'll have to get a big

[00:52:03] Kalen: boy. How do people do that? Because even if you have that kind of memory, like, okay, you're gonna run a model that's gonna chew up 80% of your memory or something like that,

[00:52:13] David: I think they usually run it on a dedicated machine and then like talk to it over the network or something.

[00:52:18] Uh,

[00:52:19] Kalen: so they're basically hosting their own server. That's the

[00:52:22] David: way to do, yeah, exactly.

[00:52:24] Kalen: Eh, that makes sense.

[00:52:27] David: It seems like, Dave Azar or whatever his name is. Yeah. It seems like he's tinkering in there. Oh, nice. So I'm kind of excited to see what he had some interesting tweets talking about that.

[00:52:36] Kalen: Nice. That's cool. Yeah, I mean, it does seem like the open source models are pretty good. I mean, yeah. I mean, I, I just feel like it's, it's so crazy to me to, to not be just using Cloud coat, but

[00:52:50] David: I know there's a couple ways it can go. It'll go the way of TV subscriptions,

[00:52:55] Kalen: right?

[00:52:55] David: Where you gotta have like five different subscriptions at all times.

[00:52:59] Right. Which is kind of where we are right now.

[00:53:00] Kalen: Right.

[00:53:01] David: Or people start like trying to host it locally after we, we gain a lot of like, incredible efficiencies in the, the inference or whatever. Right. I'm guessing we end up getting a lot of efficiencies in inference and then like all of our little, like my fridge will run AI or something, I don't know.

[00:53:22] Kalen: I always forget what inference is in the, in the process.

[00:53:26] David: I think that's just the, like generating tokens from an input.

[00:53:31] Kalen: Okay,

[00:53:33] David: let's look it up. Yeah,

[00:53:35] Kalen: it's one of those words, I sort of see it and I sort of like nod my head along and then I'm like, I don't exactly know what the heck. I know I've looked it up on a diagram at least twice, but then I've sort of like forgotten.

[00:53:49] David: But yeah, I mean, yeah, I think, uh, like the way that I'm using it is it's, you're not gonna be training a new model on your local thing. You're gonna be running one that already exists and just need to like get a response from it. Oh, okay. Gotcha, gotcha.

[00:54:02] Kalen: Dude, I mean, if, if there's good open source alternatives, that would be sweet.

[00:54:06] , It seems like people are coming out with pretty serious open source models, but I'm like, well,

[00:54:10] David: what are they? Yeah, Facebook's still working on it. Yeah. They're kind of like the. The originators of the open source models.

[00:54:17] Kalen: Yeah. Like I don't understand their whole move.

[00:54:19] 'cause obviously it's not like, like their best stuff is they're gonna always want for themselves, aren't they? Like, I don't know. I know there's the weights and then there's open weights and open source and whatever. Yeah. I don't know what they're sharing, but like, it feels

[00:54:35] David: like they've always been very, like they shared, like they invented React, right?

[00:54:39] Right. They invented, HHBM back in the day. They just always, like, they're not in the business of selling software. They're in the social network business and they just like, because of that, they've got this cool shit.

[00:54:51] Kalen: Huh. Okay. That would be, that would be cool. , If it was like really all fully, fully open, anyone could run it.

[00:55:01] yeah, dude, , good times. One last shout. , I don't know what time it is. Oh damn. It's about that time shout out was, um, Jonathan Moore, , who I think is in the Austin area. There's like three Shopify developers in the Austin area. I don't know why there's not more. It's so weird to me. , He mentioned he uses a Shopify research subagent in Claude.

[00:55:21] By the way, it's funny that all we're talking about is Claude, this whole entire thing, but, , I was seeing how the MCP was like, and I noticed it, like, does it all the time. It like, it'll return a response with like 37,000 tokens and then it shows an error message that says you, you can only get a max of 25,000.

[00:55:42] And I'm like, why the hell is it sending back so much? Um, whoops. You know what I mean? And like, anyways, so he said he uses the chauffer research subagent, and then it like, gets everything right. And I'm just like, that's crazy. I don't know how these subagent work, but there is some real magic to them.

[00:56:02] David: , Does it just, it knows like what the response is gonna be like, and so it tailors the request to make it reasonable.

[00:56:09] Kalen: I mean, it must just ha do a really good job of researching Shopify related, web stuff and returning the right answer. Mm-hmm. Because, I don't think it's hitting, I don't, subagents don't like hit CPS necessarily.

[00:56:25] David: They just, yeah.

[00:56:26] Kalen: I don't know how the hell they work.

[00:56:28] David: Um, I ran into something like that recently. Again, a Google sheet thing. Like I, I have this sheet that I. , Every new product review we get just goes in as another row in the sheet. Mm-hmm. And I wanted to like, apply a filter for a particular product and then analyze the reviews of the currently visible cells.

[00:56:50] And I, I was like, oh, that's cool. I'll just write this function that calls off the chat GPT and give it a range. But, the range still includes the rows that are filtered out of the current view. Yeah. And I, I ended up getting the like, oh, you're trying to send 50,000 tokens that's above your limit.

[00:57:10] I was like, what is happening here? It's not that many. Yeah. It's gnarly. And then Jim and I couldn't do it either. I'm so confused about certain things in my life when AI exists.

[00:57:21] Kalen: I know, man, I'll just do it. It's all very confusing. It's all very disorienting. this whole AI revolution.

[00:57:30] David: Good times. I was talking to my daughter the other day about how,, it's really interesting because when I was growing up, the internet was the new thing and like now the internet is crazy. It's everywhere. Yeah. And the same thing is kind of happening with her with ai. Like it's just, it's fledgling right now, even though it feels amazing.

[00:57:47] Yeah. Like it's brand new and I can't imagine like what she'll experience growing up. It's,

[00:57:54] Kalen: man, that is such a crazy thought. It's gonna get so crazy and I'm gonna stop saying it's gonna get weird. It's gonna be cool. Like, I really, yeah. It's gonna be cool. I really feel like there's this pessimism that's baked into us.

[00:58:10] You know, like I heard Peter Thiel do a talk on it, how people really used to be optimistic about technology and it's like every single movie technology is the bad guy and the robots try to kill us and take over. Can't we just be optimistic about cool shit without having to always, you know what I mean?

[00:58:29] Yeah. Take the whole dev devil's advocate side.

[00:58:34] David: Well, why is everyone so scared of technology? I feel like it's because the big thing that's, happened recently is social media is like the technology, and turns out they're taking all of our data and trying to sell stuff to us. Maybe that's why people are distrustful.

[00:58:49] Kalen: Yeah, that's, I guess that's true. It feels like, yeah, I guess it feels like social media is sort of like a net negative.

[00:59:00] David: I really hope that, at no point there's like, here's a free model. It's the best one that exists, but every third message you have to watch an advertisement.

[00:59:12] Kalen: Yeah. I heard this really interesting idea for monetization. I think they said what they could do is they could route you to different models depending on kinda like the nature of your question. Well, there's already routing, so it routes you to GPT five mm-hmm. Versus GPT thinking. And that you could have more sophisticated routing that, if you're on a, a free tier, maybe you get routed to the lower models more often.

[00:59:39] Or maybe if you ask a question that's related to something that could be monetized, they could not like, not change the answers, but if the answer happened to be a product. Then they could surface it with a link. You know what I mean? So that, yeah, it's not an advertising model.

[01:00:00] I don't know. It's like very close, I guess. But

[01:00:03] David: yeah, and you'd have to trust them. That might be why, like

[01:00:05] Kalen: you'd have to trust them that they were actually, but they could make that verifiable. And, and I think even, I think like if you could trust them that they were not, , impacting the accuracy of the answers, and then mm-hmm.

[01:00:20] There's just a link that has like a referral commission, like that feels like it would be amazing,

[01:00:26] David: I guess. Yeah. Kind of like Google ads,, when you search for something, Google has like a little section that's like ads that people kind of put against a certain search term.

[01:00:36] Kalen: Well, I hate those. I mean, I hate because the, because 'cause those are getting in the way of what you're trying to get to, you know, sometimes.

[01:00:47] I mean, sometimes they're relevant, but maybe most of the time, yeah. I mean, , yeah, I don't know. , It's hard to,

[01:00:58] David: you're like, I love that. Let's just hope that energy gets cheaper and computers get cheaper and there's gonna be no need to squeeze everyone for running these models. Yeah,

[01:01:08] Kalen: yeah. Hopefully.

[01:01:10] Let's hope for the best man. ' it's gonna be a sad day if it doesn't work out. I

[01:01:17] David: agree.

[01:01:18] Kalen: I'm gonna have a rough week that week.

[01:01:24] David: And then I'll get over it.

[01:01:26] Kalen: Yeah. Oh dude. I'll be hitting the vodkas hard.

[01:01:30] David: Maybe that time, maybe by then I'll be retired

[01:01:32] Kalen: and I won't need ai. I'll be slamming vodkas of the garage, dude.

[01:01:36] Try garage, get

[01:01:39] David: some garage vodka going. Try to

[01:01:41] Kalen: try to, trying to get my dude. I'll be sipping on garage moonshot dude. Trying to, trying to get my spirits up, dude.

[01:01:51] David: What would you do if you retired? Like right now, would you hang out in your garage and make moonshine?

[01:01:56] Kalen: Dude, I'd move to Costa Rica and surf all the time.

[01:02:00] David: Oh, you would surf? Yeah. Yeah, that's what I wanted. I didn't know you surfed. That's cool.

[01:02:04] Kalen: Yeah, I went there to learn to surf when I was, I guess a few years back and I was just like, my friend had just sold a company and I'm, I was texting and he was in Hawaii and we're texting back and forth, oh, what are you doing?

[01:02:14] Oh, I'm learning to surf. And about the sixth time that he texted me that I was like, dude, I wanna learn to surf., and then I was like, I was like, you know, I know what's gonna happen if I ask my wife. I know she's gonna say no, and like, she's gonna kind of like discourage me and then I'll probably not do it because like, whenever she discourages me from doing something, I just don't do it.

[01:02:36] So I was like, I'm booking the trip and I'm gonna, and, and also I had a feeling she wouldn't want to go or whatever, and I, I go, I'm booking the trip, and then I'm telling her, Hey, I'd love for you, for you guys to come if you want to. But I'm gonna go, like, I'm, I'm gonna go. And so, so that, that was what I did.

[01:02:53] That was what I did. And, , I felt very guilty, of course, just going on the whole time vacation with the wife and three kids at home. Also, the weather ended up being, that it froze out here. So they were like freaking

[01:03:07] David: out about the

[01:03:07] Kalen: weather and I was like, Hey, see you later. I'm gonna Costa Rica.

[01:03:10] Sorry. Um, and my wife. I was there for a week long trip, and then I like extended it and then like, uh, oh, no. And then like my, my wife finally eventually came down and joined me after like two weeks. And then we spent the next two weeks there together. And then we came

[01:03:29] David: back. Whoa.

[01:03:29] Kalen: Cool. And, , yeah.

[01:03:31] And so then I just fell in love with it. I was like, I wanna wake up every day for the rest of my life and like, do this. And then after that I said, okay, we're moving here. , Let's make another trip. So we made another trip for the following, like summer, a two month trip. I was like, we're gonna visit every city.

[01:03:48] We're gonna figure out which city we like best. We're gonna figure out the school, all that kind of stuff. And then we went back and then it was just like. Different. It was like the kids were fighting all the time, ah, finding activities for them to do was tricky. Me and her were arguing because of the, like, we were in too small of a place, so like, it was just crowded.

[01:04:10] Gotcha. You know what I mean? All those types of things. And then it were just like, and also it was like, well, they have their friends back home. They have their whole, all their class. Even though we homeschool, it's like, we're still very tied to this place because all their relationships and stuff like that.

[01:04:25] Yeah. So it was just like, \ it really doesn't make sense. So then I kind of let it go. But, uh, if, uh, but yeah. But if I was retired, definitely.

[01:04:34] David: What about you? So when you're, you're an empty nester, you're going back,

[01:04:36] Kalen: what's that?

[01:04:38] David: When you're an empty nester, you're going back. Oh yeah, dude. Yeah, a hundred percent.

[01:04:42] Kalen: What about you, man? Pottery store, right?

[01:04:44] David: Pottery. Yeah. Maybe I would be making pottery. And, if we're talking about somewhere that you would, that we would also just move and have a completely different life. Yeah, I'd do pottery in, Italy. Move over to, what's it called?

[01:05:01] The, the floating city. , Why can't I not answer this? Floating

[01:05:05] Kalen: city Sounds, sounds

[01:05:07] David: familiar. Uh,

[01:05:08] Kalen: Quick Google search,

[01:05:10] David: quick Google search. Venice. Hello, Venice. Ven, Venice, Italy. I would go to Venice, Italy, and I would make pottery and sell it to local old coffee shops. Hell

[01:05:20] Kalen: yeah,

[01:05:21] David: dude.

[01:05:22] Kalen: Damn, dude. How cool would that be, man?

[01:05:25] If let's say we like lose touch 10 years from now, 20 years from now, we run into each other in Italy, you've done exactly that. You know what I mean? I'm like buying pottery from your store, talking about the good old days. Heck yeah.

[01:05:40] David: Like,

[01:05:41] Kalen: that'd be rad.

[01:05:42] David: Let's get it going. Let's set up the plan.

[01:05:44] We got, Let's see, I have like 10 years maybe 12, 15. Let's call it 15 years until my kids are out of the house.

[01:05:52] Kalen: Yeah.

[01:05:53] David: So gotta start saving,

[01:05:55] Kalen: start saving up that dough, man. Um, start saving up that Bitcoin.

[01:06:00] Good.