Bản chép lời đầy đủ của John Nunemaker - The Conductor

Smith (04:36.95) Yeah. Yeah, I can see that about you. Mm-hmm. That makes sense. Jeremy Smith (04:53.422) yeah, yeah. That's really key though, it's like being able to work with your energy level, because it can really unlock things rather than just fighting through other people's priorities, just kind of powering through a bunch of work that is kind of against your nature where you are. at or where your what your curiosity or excitement is around. And not everybody has that opportunity, but being able to get to a place in your career where you can do that and ride the waves of your own excitement feels like that is something we all should be shooting for. John (05:23.158) Mm-hmm. John (05:36.63) Yeah, it's been it's not easy. would say it's been a very intentional thing like my entire, you know, the time I've been working. So because I started with that kind of from the beginning that like every chance I got, you know, that's where I kind of shifted things. So I never tried to like make the most money or things like that. I always tried to increase the flexibility. so, you know, doing that means that I can do this now. And it also means I can take, you know, two month camping trips in the summer and things like that. And I think, you know, sometimes I do hear it where people are like, I imagine how can you even do that? And it's like, well, if you've spent 20 years working towards something, then it's possible. didn't do it, you know, for the first 10 or 15. So yeah, it just takes a lot of time. And it's the same thing with being able to mentally jump between those things, you know, and it's, I just believe really bigly in largely, hugely, bigly seems the wrong word. Yeah, that's the wrong word. Bigly. Like I just, I really believe in fractional and like Jeremy Smith (06:25.464) Bigly. Bigly's good. Nope, I like it. Jess (06:25.863) I like Bigly. John (06:35.255) putting time in lots of things and don't put all your eggs in one basket. I think. Exactly. Yeah. Jeremy Smith (06:39.588) man, you're preaching to the choir here. That is everything I'm about. Yeah. That's cool. Jess (06:48.246) and you forgot about the offline business, the real estate business. John (06:57.237) Yeah. Yep. Yeah, I forgot about that. So momentum. is one another thing that I've been involved in. so it's a entrepreneurship hub. So coworking office space. think we have like a little under 50 offices and a very big project. I'm I was like kind of the first non-founder whatever investor person in it because one of my friends is is one of the the main people. And then and it's got like a podcast studio. It's got all kinds of stuff. It's really cool. So we end up I'm not in there today. I probably should be because then I would sound better. I'd be on a sure SMB seven with, you know, Jeremy Smith (07:29.792) Yeah. John (07:30.616) a good setup instead of my Yeti Blue mic at home. I mean, I am at least rocking a boom arm. I'm not like, you know, sitting in on the

heart or whatever surface, but, yeah. So I, that's been really cool. It's been fun to see kind of fill up. think we're maybe halfway through the offices, maybe more. haven't really looked at it in a while, but it definitely feels busy. have, you know, awesome coffee draft lattes, you know, all that kind of stuff. It's really, from a local coffee company. That's really good. Jeremy Smith (07:36.332) Yeah. Jeremy Smith (07:53.731) man. John (07:58.666) It's been awesome. like, honestly, it's been great because of the, is becoming a community. Like that's been the best part of it. It's like South Bend's kind of small, like we're a hundred, 150,000 people. And so getting like, to the point of having. Jeremy Smith (07:59.0) Let's go. Jeremy Smith (08:03.999) Mm-hmm. Yeah. John (08:13.448) a community and having like like being around other people that are programming and stuff like that. I've just never had that here. It's usually just the people I'm working with and they might be spread all over the place. So that's been really cool to just talk about like A.I. in the morning with, you know, Matt or JP or things like that and be able to show people in meat space like, this is how I do it. That's been that's been really fun. So so, yeah, it's I think it's going well again. I haven't seen the financials or any of that stuff, but it's one of those things where it's like if you don't if you believe in something Jeremy Smith (08:32.523) Yeah, that's cool. John (08:43.264) then you put your money behind it, even if you don't know what's going to happen. And so that's, you know, I'm like, want that to exist. So I'm going to invest in it. I'm going to show up. I'm going to, you know, recruit people to be there. mean, I probably five or six of the offices are people that I pitched and toured and all that kind of stuff. So yeah, it's been, that's been really exciting. Jeremy Smith (09:01.943) That's Jess (09:05.23) Does that co-working space have a basketball court in it? John (09:08.872) man that's it's missing that it has a full gym so it's got a full like you know treadmills and all that's and free weights and kettlebells and machines and all that it's it's really well you know done well there and two shower full full like separate bathroom showers and like tiled they're not just like we didn't throw in a surround they're they're they're nice but yeah no but no vast well hope i have to go home for that so Jeremy Smith (09:09.613) Yeah. Jess (09:10.254) You Jess (09:19.425) wow. Jeremy Smith (09:24.298) yeah. Jeremy Smith (09:33.727) Yeah. That's awesome. Jess (09:35.47) play in the barn. John (09:36.703) Yep, that's right. Jess (09:40.569) Well, just maybe one or two follow-up questions on current businesses. what, so people have heard me talk about buying businesses here before and has there been, was it, was it what you thought it was or was there any big like, this is, this was really a gotcha like that I didn't recognize or Is it the

things that people say it was like, okay, you're buying a system, you're buying customers, you're buying a platform, a team, processes that are already in place. Was it all that? John (10:20.895) Yeah, I mean, we what we did, I can think it was a little bit unique and I actually love it because we didn't buy a team. We basically just bought, you know, a bunch of servers and domains and then obviously customers a lot that are using those. So that I think made it a. yeah, fire, fire side. Yep, yeah, there's too many sides. It's there. Then a lot of eyes and ease. No, you have no so many people. You are a Riverside is right there in front of us. So I. Jeremy Smith (10:34.625) We're talking about Riverside here. sorry, Fireside. Jeez, I did it. I did it, I did it. Well, I'm also looking at Riverside right now, so yeah. Yeah, whoops. John (10:49.235) I've done it before too and lots of other people. so I think with yeah, with fireside, was like that made it a little bit more simple because you're not like trying to change people's bring people along and change their culture and all that kind of stuff. It was just buying like the tech stuff. So from that standpoint, I would say it's been everything I expected it to be. It's also been what I expected from like, you know, I would say more technical debt and more fixing of things than I was initially hoping for. Jeremy Smith (11:16.791) Mm-hmm. Yeah. John (11:19.477) But that could just be that's the kind of stuff I like to. So I gravitate to that. And I think a lot of it has been support driven. Every time I get a support request, I'm like, OK, like, what can I do to make this a good experience and make people want to stay and things like that. So that can definitely lead you. Jeremy Smith (11:35.566) That was one of the things I was thinking about when I was thinking about the different types of work that you might be doing. seemed like with fireside, you might be able to do more of the support driven stuff where you can make this customer's day and you don't have to spend a week shipping a brand new feature or thinking through the architectural changes or something. It's like, we can fix this one thing. And now you can do that even faster using like agents and stuff like that. So you might be able to unlock really much faster. John (11:40.424) Mm-hmm. John (11:52.552) Yeah. Jeremy Smith (12:03.541) And it's a different kind of thing than having to spend like the entire day like figuring out the architecture of this thing and having discussions and bringing in a designer and all that other stuff. John (12:08.5) Mm-hmm. John (12:12.584) That's exactly it. mean, right now I get, you know, I got a support request. I was like, you know, my player is not responsive enough. And I was like, okay. And so I just made a new responsive player, like from scratch in like three days while I was doing other stuff and I shipped it and there was a couple of bugs, fixed them. And it was out and it good. And then another customer was like, I would love play lists, you know, like not just an embedded player that's responsive, but I want to have like a list of them. And I was like, okay, like, so I made, just, Jeremy Smith (12:25.089) Hmm. Yeah. John (12:42.265) made that and it was super easy and now you know show your latest however many you know you want to show and then another customer was like well I want to have curated playlists like I want to pick things and I was like sure you know so like that's that's been done for like three weeks I've just had other stuff I have to like actually merge it and deploy it. Jeremy Smith (12:54.252) Yeah. Jeremy Smith (12:58.263) That's cool. John (13:01.159) And then also because we're adding the new studio, like that's been a one thing that kind of slows things down because, know, we're doing two at once so that people can get comfortable with it and they can opt in. And then eventually it'll be kind of a forest opt in once we feel pretty good about it. And, you know, the great thing about that is we can do that. The downside is then everything new has got to be built twice. So you kind of freeze for a little while. So that's, that's been the freeze right now. So I'm working on migrating that from the existing, studio into the new studio. And then, you people will have it they Jeremy Smith (13:05.377) That's a bigger. Yeah. Jeremy Smith (13:20.781) Yeah. Yeah. Yeah. John (13:31.035) or whatever they want to do for that kind of stuff. that's, and one person was like, I want to save my customized color, you know, like, and I was like, done. they're like, my gosh, like I just asked for this and it's done an hour later, you know, like, that's been, yeah, it's been amazing to just kind of pipe, you know, I don't, Jeremy Smith (13:46.783) Right. That's great. John (13:52.483) actually, you know, pipe it straight and have like an attack vector or anything, but I'll, you know, copy and paste the key things and even images and stuff like that. And then just have Claude work on it. My best friend Claude. Jess (14:02.989) No, he's my best friend. Jeremy Smith (14:03.018) Yeah. That's funny. Yeah, that's really interesting and it like speaks to different kinds of work that you're still getting done as a developer. Cause some of those things do take the integration time or the time of like transitioning customers on existing tools. But then these smaller bug fixes that give you these quick wins that probably turn into like customers that end up raving about you. John (14:06.184) That's not what he says. Jeremy Smith (14:34.3) And some of that feels like maybe it's like so different from the experience maybe they've had in the past few years. Like when a product has been sitting dormant mostly, it's still kind of running, but that helps people feel like, okay, this is moving in a new direction or it's gonna stay up with, you're making like kind of new commitments, I guess, to the product, which gets spelled

out there. That's cool. That's awesome. John (15:00.272) Yeah. And they've been raving. I mean, it's been, that's been the most fun thing is like with box out, I'm, I'm a couple layers away from support. Jeremy Smith (15:08.81) Yeah. John (15:09.072) Like, I mean, the thing I'm doing right now for box out, one of the things is like, literally, I'm, I'm, I'm slurping in all of our support requests into the database so that I can, you know, and creating an API on it so that I can MCP it so that I can like, just chat with our support requests to be like, what's the most common over the last week or like, has anybody had this issue before? Like things like that, because I'm like, I'm just out of touch. I don't, I don't, I'm not in there. have a full time person. Jeremy Smith (15:17.834) Mm-hmm. John (15:39.075) who does that and if they can't answer it there's you know other people in between before it even gets to me unless it's like a really weird esoteric technical thing so for the most part Jeremy Smith (15:50.731) And does that mean that the support volume, because the products, more mature has higher usage, you've got a higher support volume then that makes it worth it where if it was a smaller product, you could probably still scan through support or kind of more easily. Yeah, okay. John (16:04.103) Yeah, you could. Yeah. So we're like, don't know where it's exactly at. mean, I would say for the size of revenue, it's probably a small amount of support requests. We automate a lot of stuff, but it's a couple hundred conversations a month and probably each one of them is, you know, could be 10 to 30 messages back and forth. Um, I mean, a lot of the support requests are just like, help me find a good graphic. Like that, this is what I want to do, you know, which again, that's probably stuff we, we, I've been thinking we need to just solve with AI and with inspiration, you know, show what, Jeremy Smith (16:34.625) Mm-hmm. John (16:36.602) what other people are doing and stuff like that. So that stuff's on the roadmap. But this one, I was just like, I really want all the support requests somewhere. And then eventually I want to vectorize them so that I can do semantic search and all that stuff. Then I've been setting up this whole company brain thing for box out where it's like, you know, LibreChat, which is like an open source chat to BT. and with MCPs for like our read replica database for Stripe for, now I'm going to do another one, you know, just homegrown for this internal API. So I'll have internal APIs. Jeremy Smith (16:38.41) Yeah. John (17:06.679) on in our app. and then MCP that talks to those internal APIs and then that gets installed in LibreChat. now sales marketing, like everybody else can talk to, you know, can look at stuff the same way that I do. And I can set up recurring crons in a easy place and have it, you know, federate out to everybody. so that's, that's, kind of just came from like, I was like, well, I want this

like company brain somewhere that has everything. And so I was like, well, I'll just spin up a server and start throwing some stuff on it. And then I was like, this is actually Jeremy Smith (17:31.232) Yeah. Yeah. John (17:38.039) kind of cool like I was like how many graphics do we have in it it counted them and did the figure out sequel query from our database and showed the number so yeah stuff like that I was like magic well so when a tiny little integer loves another little integer they have vector math and that no so how it does is there's a postgres yes that's basically how it works a special forever hug so basically the Jess (17:47.615) How does it do that? Jeremy Smith (17:50.764) You Jess (17:51.981) I mean, do you have to build a tool for it to... Jeremy Smith (17:55.358) You That's how that works. John (18:07.78) you can beep that out. there's an MCP that's the Postgres MCP for, know, just like if you just search Postgres MCP, there's a few of them. So you get that you can, for LibreChat, it's just an open source, you know, Dockerized thing. So you just run it with Docker wherever you want, it doesn't matter. And so you just configure it and say these MCPs with these, you know, environment variables and all that kind of stuff. And then, so that's how... Jess (18:31.096) So LibreChat knows about the Postgres MCP server. You like config that. then, so when you ask it a question, can, I guess it can see your schema and it just automatically knows how to run the query. John (18:34.906) Yep. And then the post-crisis. John (18:39.76) Yeah. Yes. Yes. So it'll infer. And again, it's not going to be perfect. So like one, the first one thing I asked it was like, how many unique social accounts do we have? And it was like, it was like a eight, 8,000 something. And I was like, we don't even, I don't know if we even have 8,000. We don't have 8,000 active orgs. We have 8,000 orgs, but like not active. So I was like, that's gotta be teams, not orgs. So then I was like, okay, but like join orgs. Jeremy Smith (18:49.13) That's. Jeremy Smith (19:03.681) Mm-hmm. John (19:10.843) I said, is that orgs or teams? And it said teams. And I was like, but I asked you for orgs. So join orgs and then tell me unique. And it was like in the two thousands. And I was like, okay, that makes that's more sense. And I'm like, one of our salespeople is not going to able to do that, but they're going to be able to probe it and then be like, you know, mentally do some math to say, does that seem right or not? And especially if I help, you know, help them and teach them this kind of stuff. And then they can come to me and be like, how do I make it more accurate? And then I can say, I can see You know, the, things that are not quite working and I can add that. So Lieberchat has an idea of custom agents. And so they come with like a default prompt context type thing. And so that I can make a database agent that automatically uses the MCP, like enables it and all that. And then in there, I can say like, this is how you get active users. This is how you get active or I can put some of the common things in there so that it can kind of more, correctly pull the data based on when people ask questions. That's my thought. I haven't done that part yet, but I'm I'm 99 % sure that that'll get most of our use cases. It's pretty simple things usually. Jeremy Smith (20:13.545) I've spent enough time doing like, in the past, doing like, setting up Metabase and then writing custom queries for the C-suite to figure, you know, to understand what's happening with customers that I know that like, it's been tricky for me to translate their questions into SQL. And so this is where I get to be like, I don't, like, I don't totally trust this. Like, I've been wrong about things where it's like, the C-suite wants to know, how many customers were active last year? And it's like, what do you mean by active? Like, what do you even mean by active? You've got to define what that means. Like, what do we have in our database that means activity? I don't know what you're, and so we have to have that full discussion, which often they don't even know, right? So then you got to figure that out and come down to like something that ends up being a SQL query that you are very confident is providing the right answers and. John (20:46.521) Yes. Yep. Agreed. Jeremy Smith (21:08.981) is not including soft deleted records and all kinds of other things that you might not be thinking about when you're first writing that query. And so this is the place where I totally get that the speed thing is super beneficial in the accessibility, maybe the speed and accessibility, but the accuracy ends up being this big question mark to me, because it could lead people to make completely wrong decisions if they weren't, if, John (21:11.971) Yeah. Jeremy Smith (21:37.963) if it was invalid, know, like completely wrong business decisions could be made. so, like, what are you, what are you thinking about that? John (21:43.051) Yes. John (21:46.572) Yeah. So how I'm thinking about that is like, Hey, we have a small team. So, you know, I'm talking about two, three people that are, are going to use this, that are not technical, maybe four or five. So that reduced the blast radius. I can more easily kind of like check, you know, on what they're doing and whether it's accurate and stuff like that. But the biggest thing I'm thinking is like, we also have Metabase set up and I have a bunch of stuff in there. And so generally like I, I would have them, I would put stuff in there, but I think one of the things that's kind of cool that this enables is like in my head, I know Metabase has an API. Jeremy Smith (21:49.983) Mm-hmm. Mm-hmm. Jeremy Smith (22:01.514) Mm-hmm. John (22:19.782) And so my plan, haven't made it this far yet, but my plan is basically to use the Metabase API, probably make an MCP, small MCP for it or something, then make it so that, or

skill, whatever, so that that will maybe get hit first. So if you say like, how many active subscribers do we have? Well, we've got a report that directly matches that by name. So it can pull the more accurate stuff, you know, using that. But what I want is I want, you know, the cross, I don't know, pollination. Jeremy Smith (22:35.616) Mm-hmm. Jeremy Smith (22:41.161) Yeah. Jeremy Smith (22:44.459) That makes sense. John (22:49.476) I have the Stripe data, I end up with the customer support requests, I have subscriber data, I've got the raw database, and then I want the LLM to be able to work across all of them to kind of figure things out. And so I think that's the thing that I think is the most valuable from it. And my plan is basically to, the things that I see, that people want is just to make it available in one place. And I think the key thing is really just, you know, whatever, like an open source version that you can configure, you can control, and then everybody's not installing these random keys in your database passwords and stuff on all their machines so that they can just ask a few things, you know? I mean, everybody, all of our team has Claude accounts and ChatGPT accounts and all that kind of stuff anyway. like, but I'm like, I don't want them adding, if I have a salesperson adding an MCP to Claude, I don't I don't think that's a great idea. Like I would rather I manage it, configure it, and then I help them to get more and more of the stuff automated. So I'm kind of sales enablement in that way rather than the opposite. And I think like the biggest use case, what was kicked it off for me is like you get attendee lists from conferences, you know, when you're a vendor. And so we had a person who was going through them and he went through like a hundred out of like 400 and it took him three hours because he's going into our system, pasting in the email view, view subscription. Jeremy Smith (23:42.239) Yeah, yeah. John (24:10.526) Okay, you know, or no, they didn't show up. So they don't exist or they showed up, but they are no longer a customer than, you know, not current and all that kind of stuff. And he took, it took him three hours and I did it in like five, 10 minutes with a rake task. And I was like, this should be something where, you know, if I make an internal API, that's like hydrate this email. and expose it as an MCP to something, they should be able to upload a spreadsheet and then hydrate it from, you know, from AI somewhere. So that's kind of what, kicked it off is like, let's start with that use case. And so now I'm just kind of slowly adding data sources to try and help with that hydration. And then I fully expect I'm going to have to make very specific reports. Jeremy Smith (24:53.291) Mm-hmm. John (24:53.667) And pull meta bit is already on my to-do list to pull meta base Into this equation as well because we have a bunch of stuff there. So that's how I'm approaching it right or wrong Jeremy Smith (25:00.907) That makes sense. Jeremy Smith (25:04.627) It feels like there's a missing verification step and maybe it's maybe what you're saying is you're like you're saying like on a small team, you are the one that's kind of like spot checking. But I think it's still I still feel nervous about this for a good like here's a good example. I had a couple of weeks ago. We were announcing the speaker lineup for Blue Ridge Ruby. I was so busy like trying to wrap things up that week. I you spun up conductor. John (25:15.337) I am. Yeah. Jeremy Smith (25:34.205) set up the Bridgetown site for that's what we're using for the site. And I said, like, here's what the speaker pages have looked like in the past for past years. Here's the selected speakers, just build out all the pages and go like find all of their social media profiles online and add those. You know, add their LinkedIn, add their Twitter, whatever else, right? It got really far. John (25:48.086) Yeah. Yup. Jeremy Smith (26:01.662) Like even pulled in their profile photos. I had it pull in their profile photos. It got really far. would have, it was so tedious. It would have taken me forever to do this, right? But then I come back through and I'm looking more closely and I find that like for one of the speakers, it pulled a different Twitter profile using the same handle that the person often uses. But this profile was not, not safe for work. John (26:01.899) Yeah. John (26:22.71) Mm-hmm. John (26:30.814) Yeah. Yeah. Jeremy Smith (26:30.954) It would have been really bad like if I'd like put it would have been understandable Like I could have scanned that and been like, yeah, that's their handle. Nope. Not on that not on that service so That that would have been really bad. It would have been really bad if I had just been like, okay good We're good to go. Also it like copied over bio information and like change some things It was right and some of the things that changed like fix some spelling problems but in other places it like massage like what was being said and I didn't realize it. So I had to go back over each one of them and like recopy the abstract and bio again because I didn't trust it. So this is the thing that's like without like something else that comes in behind to ensure like that you don't get these surprises and like I could have just scanned through it would have looked fine. That's the thing that's like to me really scary. John (27:23.041) Yeah. So what, like what I would, how many speakers like 40, 13, 15, 10. Yeah. So like what I would do is that. Yeah. I would have did like a, I like what I've been thinking a lot lately is like, how do you verify, know, like, cause if you give the LLM a way to verify. Yeah. Jeremy Smith (27:26.718) It was 10. No, it was only 10. Yeah. It still would have taken me a couple hours. Jess (27:38.07) Yeah, that's everybody's mind. John (27:41.492) So in this case for that one, what I would have did is I would have done like a

couple of phases. Like first phase would have been like, you know, gather all the profiles that you think might be correct for these people. And then, you know, like you could kind of spot check them and like your yourself. And then once you feel like you know, that's pretty good. Then I would probably Jeremy Smith (27:54.952) Mm-hmm. Yeah. John (28:01.438) say given these profiles like, you know, make the pages and stuff again, it still involves a human in the loop. Maybe it's a little bit easier and fewer tokens or something in two phases than in than in one. There's definitely, you know, again, it's like, if you had like this magical prompt, you know, you could definitely have a magical prompt that was like, make sure these are always not safe for work and make sure this and make sure that like you could do that. But like for this task, then you're going to spend more time. And it's like, how many times are you going to do this once a year? Maybe, you know, so it's not worth it. So I think it's okay to have it do that once in a while. And then if it's something that happens a lot, then you do need that verification process. Like you need a better prompt. You need to like, Jeremy Smith (28:23.338) Mm-hmm. Jeremy Smith (28:29.938) Right. Yeah. John (28:41.343) you know, or a skill, whatever, something where you can actually, you know, get a loop going that's, you know, virtuous that will get the answer a little better. I don't know. I, it, I, same thing happens to me on a lot of stuff. it just doesn't really bother me because I still feel like overall it saves me time and I plan on being the human in the loop and anyone that I'm, that I'm showing this or teaching this, I try to explain that to them as well. is like, look, it's not, it's not perfect. And even it, it's only, it's just as good as a human. could have sent a human after that and they could have got the wrong one, you know, like in the end, it's all about the goals. And so you, you knew like, well, when I look at these, I'm going to make sure that like, this is exactly accurate and I'm going to make sure that, know, but the nice thing is like a lot of the URLs are right there. You can click them and immediately just, you know, scope it out and kind of get a good sense. yeah, yeah, it's tricky. Jeremy Smith (29:31.583) One of the things that I hated about that though is like if I had given a person like sent like an intern to go to if I had a Bloor's Ruby intern, you know, and they made that mistake I'd be like this is a big mistake, right? I need you to understand this like if we'd gone live with that that would not have been good and they'd be like, shoot, you're right I'm never gonna let that happen again, but like it's Yeah, right. That's true. I guess it will say that I will not believe it John (29:39.872) Yeah. Jess (29:53.528) Claude will say that too. John (29:56.491) but you could put that in your Bridgetown Clod. Yeah, but you could put that in your bridge down, Claudette, and I mean, MD file, like, if I ask you to like research speakers, make sure that they are safe for work. And if they're not, give me a warning, you know, like, that's, that's what I've lately I've been doing a lot, actually, is just putting stuff in Claude MD. Like every time something is wrong, I try to put it into Claude MD and be like, that way, it's not wrong again, you know, Jeremy Smith (30:03.594) Yeah. Jeremy Smith (30:07.72) Yeah. Jeremy Smith (30:14.644) Mm-hmm. Jeremy Smith (30:21.118) Yeah, yeah. John (30:21.822) The only thing that's weird is, you know, if you work on a team and some other people put things in, then it's like, no, like that's not what I want. so some of that is a little bit interesting. Jeremy Smith (30:24.958) Yes. Yes. Right. That feels like a very weird. Yeah. You just highlighted a thing that's really kind of drive me crazy. I don't know where to put my stuff that I want to tell Claude that I don't necessarily want to mess with any, like any of my clients, you know, the way they have things set up. I don't want to mess with their Claude setup or their agent files or whatever they're doing. I just want to have my way to do it so I can direct things. So I've been doing like the local Claude, like local MD. and I said like, John (30:52.074) Yeah. John (30:56.736) Mm-hmm. Jeremy Smith (30:58.234) set up a way to do like a it just pushes to a gist a secret gist so i keep that synced that way but it's not great it's not not a great situation John (31:10.718) Yeah, for the most part, try to keep, go ahead, Jess, what were you gonna say? Jess (31:13.976) I was just, have you guys used Brian Castle's agent OS? Jeremy Smith (31:20.06) Haven't no, think you were you were have you have you done that? Jess (31:24.214) I haven't done it, but we had him come to a workshop for our team and he kind of gave us an overview of that. And one thing that's kind of cool is that you can, it's sort of my best way to describe it is that you like create a Claude MD, but for specific things, like you can have one for like models or services or different ways. And the cool thing is, that Claude doesn't reference it unless you tell it to you. So. Jeremy Smith (31:25.16) Okay. Mm-hmm. Jess (31:51.727) You know, if you're doing something and you don't want to inject it with that context, then you don't have to. But if you know you are, then you could just say like a certain, I think, slash command or whatever, and it pulls that in so that you can use that history that you've already documented for like a certain way that you create a service object or whatever. Jeremy Smith (31:57.995) Mm-hmm. Jeremy Smith (32:15.934) Yeah, I think that's cool. It still feels like whatever we're using, maybe needs to be coordinated at the team level. I don't know. Like I think that's where it's weird. Like it feels like it's right. This, it's not like, linters where we're picking our Rubo cop rules or something for the team. Maybe

it's like that. I don't know, but it feels like maybe so, I guess. Jess (32:38.754) It's similar, I see. I think. Jeremy Smith (32:43.058) I guess because it still feels experimental, it's hard to put as like these hard and fast, like we're committing this, everyone's going to now use this or here are the skills that we're all adding. And, at this point, I haven't been able to, I haven't felt like it was worth trying to set up that coordination level with clients to make sure we're all on the same page and make agreements about what gets included. And so maybe just cause it feels experimental still. I don't know. John (33:11.529) I also run really vanilla too, honestly. I don't think there's a whole lot of benefit in a ton of plugins and skills and things like that. I've used a few, Frontend Design Skill's pretty good. I definitely will use that. I actually installed a Rails Upgrade Skill today and I'm finding some value in that. But like... I generally I try not to use them a ton because I don't know. Everybody's different. Like I think Garrett's actually who I work with as went kind of the abstract. He's like very it's like you use it for like whatever whatever you do and however you work it accentuates that. So like you know he creates a lot of markdown files and a lot of like you know context and specs and like things like that. And I'm like I'm just like a little. Chihuahua, you know, jump in between things. so like, I'm like, don't have time to like do this big doc and do all these things are like, I'm never going to read this again. And then it's going to get out of date. And then if it gets out of date, I'm not going to like, it's funny because like he set up this whole like, Jeremy Smith (33:55.939) Yeah. Yeah. John (34:08.559) he has like an entire repo. I've been giving him hard time about it. Like all these things and he's like, yeah. Like whenever I want to update a dependency, I just run my dependency update skill and here's all the stuff it does. And he's like, just install, you know, this thing with like 80 skills in it. And I'm like, no, I'm not going to do that. Like I really, he's like, no, well I, I created them all like they're vetted and I'm like, well they're vetted by you. It might not be what I want to do. And plus like every time I've installed something wholesale with like a ton of functionality, I use zero bits of it. Jeremy Smith (34:22.843) Yeah. Jeremy Smith (34:30.215) Yeah. John (34:36.734) Like I just, if it doesn't get ingrained in me, it just sits there and it does nothing. like that's the thing that I've just been very slow, very vanilla. I try to keep to like Claude.MD maybe a couple skills if I'm just like, and if they don't, if they don't provide value right away, I just remove them because I just don't, I don't know. And a lot of times for me, I'm not even like, I don't even ref like. you know, you can add and then you can reference a file. I'm like, figure it out. Like just figure out what the

file is. I don't care if I, I saw, you know, the open cloud guy tweeted the most hilarious thing when he was like a lion doesn't concern himself with tokens. And I'm like, that's, that's, I'm not going to, sure. I'll save some tokens if I reference the file, but I'm like, I'll lose some neurons or something. I don't know. I just, yeah, I just, I go vanilla. I don't even correct things. I, the whole time all I'm doing is just talking, whisper flow, let it roll. Jeremy Smith (35:03.069) Yeah. Yeah. John (35:28.912) and see what happens. And if it doesn't figure it out, I do it again. If it messes it up really bad, I just get checkout or archive the workspace and start over. Yeah, I don't know. So I'm the least efficient AI user ever from the standpoint of process. But I feel like the end results, a lot of stuff is getting shipped. So I'm not unhappy with that. So I don't know who's right or wrong in that war. It's tabs versus spaces, probably. Jeremy Smith (35:52.49) I think what you're speaking to, you and Gara are both highly experienced developers and have shipped plenty of good work, but you have different working styles. So using your tools, it's accentuating the working styles that you've already developed, probably, it would be my guess, and sort of the propensity you have. I think it shouldn't look the same. John (36:04.839) Yep. Yeah. John (36:12.689) Yeah. Yeah. Jeremy Smith (36:19.591) Maybe there are certain standards of quality that we all, you know, aligned to and like, you know, so linting is part of that and also maybe style, you know, style, having style guides, I think makes sense, but like how you get the work done up to that point probably shouldn't be dictated, at the team level, I think. And this is where it feels like it crosses over, with what gets committed. I don't know. That's my current feeling. It feels like a, an area of contention still. John (36:46.853) I think it's even true on the models that you're using and stuff like that. Like some people are like, my gosh, Codex is just way better. And then other people are like Claude, you know, they're right. And they're like, Claude is way better. And my thing is just like, I think certain people think different ways and therefore certain LLMs get trained differently. And so people just resonate or speak better to one or the other. I don't think one is better. They're just, yeah, exactly. It's the same. Jeremy Smith (37:08.553) It's like working with people too. Like you can hear the same thing, like, this person was great to work with and the same person, you know, might for somebody else might not be. And, yeah, that's speaks more to the, the interplay, you know, inner dynamics between. Yeah. John (37:15.238) Yeah. John (37:24.56) Yeah, exactly. Like open code. Sometimes I've used it and I'm just like, it's cool experience. It's fancy. It's got like node neon colors and all that kind of stuff. But it like, we'll just randomly change single quotes, double quotes.

And I'm like, I don't care about that, man. I don't want like a dirty diff. I just, I don't, I'm not here for that. You didn't even, you didn't even need to touch this file. Like what made you decide to do that one thing? You know, Claude doesn't do that for me, you know, and Codex is just, I don't know when you see it in the output in the terminal, just, Jess (37:33.72) you Jeremy Smith (37:39.911) Yeah. Yeah. Jess (37:42.638) You Jeremy Smith (37:46.033) Yeah. John (37:54.276) feels harder to read and harder to parse and harder to gather what it's doing. know, so clock code or conductor. just I just resonate with those more. Sometimes I use codecs to like review and I do I think it's good. think I sometimes I think I get a better output. I saw the best. The best thing I saw was like a meme and there was somebody was like. Jeremy Smith (37:55.997) Yeah. Jeremy Smith (38:02.567) Mm-hmm. John (38:13.904) They said something to the effect that like Claude is like, you know, like the cool coworker you love to hang out with who slings a bunch of stuff. But like sometimes you're like, no bro, like not that. And like Codex is like the guy in the corner who's or he was like shipping a ton of things and he's amazing, but nobody talks to him because like he's a nerd. And I was like that. I kind of, yeah, I actually kind of vibe with that. Like Claude's kind of a little crazy sometimes, but it keeps, you know, keeps me interested. So. Jeremy Smith (38:22.153) You Jeremy Smith (38:28.041) Yeah, I kind of get that. Jeremy Smith (38:37.447) Yeah. Have you tried Kimi K 2.5? I think that's what it's called. I tried that last week. Did not go well. I was not happy with that. John (38:47.713) So I tried, it's another model. It's like supposedly Opus level smart. I think it's out of China maybe. Kimi and Minimax are the two I think that are both maybe from overseas. yeah, they have their own subscriptions you can use too. Jess (38:47.927) What is that? Jeremy Smith (38:50.653) Yeah. Jeremy Smith (38:54.568) Mm-hmm. Jess (39:00.013) how do you use those? Through open code? Jeremy Smith (39:00.327) Yeah. Through open code. Yeah. They right. Open code has a, like a subscription. Yeah. Zen model now where you can get. So I got, I spent 20 bucks to get credits for that. And, just to give it a spin. Cause people were talking about how great it was and I did not, did not enjoy using it. so I mean, I'll try it again, but first pass, not, not great for me. John (39:10.425) Zen? Yeah. Jess (39:12.631) Yeah. John (39:29.37) Here's what I, what's cool is open code Zen right now has both of those for free. So like if you pick, they have a, version of it that's free. And so you can, you can just use those. You don't have to even, you know, pay anything. So I was kicking the tires with that on the, box out brain project. Cause I was like, well, I want like a web version for like, from my phone where I can just type something and do something. And it does the

work trees and does all the right things and stuff. And I'm like, nothing like that exists. I don't care what anybody says. Like I have not found that yet. If it does exist, I would love to hear that. but I'm already just make it, but Jeremy Smith (39:36.903) Mm-hmm. John (39:59.363) open code has a web. You can say open code web and it'll serve on a port and then I can tail scale, you know, the whole thing and I can go to a browser and do it, which is cool. But then, you know, you have to, you basically lose your cloud subscription. You can't really use your clogs subscription. and so I've toyed with like just getting like, what I ended up doing is like, got a cloud subscription just for it. But in the meantime, I was like, well, I'll just try these free models, you know? So I did. but then I realized that they're actually going to those services and they're not like, they're not hosting them. Jeremy Smith (40:13.864) Okay. Jeremy Smith (40:22.92) Mm-hmm. John (40:29.213) because they're open source you can host them but open codes not doing that they're just routing their open router for their own things and I was like I don't actually like that because like the you know it's going wherever it's going and then what's it going to be used for so so I tried it a little bit and then I was like nah I don't think so again I Jeremy Smith (40:41.062) Right. Mm. John (40:49.487) I think the main thing is like, you doing great? Like then just ignore the noise. Like ignore what everybody else is saying. Don't feel like FOMO. Like every, all it is, like X is just this hype. cycle of like people will post like Kimmy 2.5 is amazing. And then somebody's like, Whoa, that got a lot of traction. I better post the same thing and mix it up a little bit. Claude, can you make this tweet my own and post it? So like you end up with this hype cycle of like the same stuff. But I'm like, they're all about at maximum. They're all about the same smartness. So just pick one and get your job done. You know, like who cares? It doesn't matter. So that's kind of what I've been doing. Although I will admit on the side, Jeremy Smith (41:08.774) Yeah. John (41:29.658) I read this post that was like how to write, like pick the best hardware to run the models locally. And it was like for like three to five K you can buy this machine that has like 256 gigs of Ram and can run these models like on the machine. And I was like, that actually doesn't seem like a bad idea. like because then you get local instead of remote. So like it's never down. It's always long as your servers plugged in, you're good, which is cool. Um, and I also think the benefit of that is like the prices of Ram and the price of all these things are only going to continue going up. If you believe that AI is going to continue doing what it's doing and unless we find new ways, like again, like I don't know you guys have seen chat, Jimmy, have you seen that yet? Oh my gosh, it's insane. So somebody took like

the like llama. I don't know. Jeremy Smith (42:09.618) No. John (42:14.968) I'm not even saying the right words, 3.1 or 8 bit or some, some kind of things like a cut. One of the kind of dumb hallucinating models and they wired it. Like they literally created it into like the CPU or the GPU. Sorry. Saying the wrong words. And so like it will literally sling. like the fastest one was like service or something like that was like the fastest tokens per second. And it like, dwarfed it. It's insane. So if you go to like chat, Jimmy.com right now and you just type in anything like questions, like you would type into chat, Jimmy T it, it responds. Jeremy Smith (42:27.121) Okay. John (42:44.844) fast that it feels like it responded before you hit enter. It's it's there's no thinking. It's insane how fast it is. So the future is going to be that you know, so which is awesome. And eventually that will be like you could run it locally too. So there's a part of me that wants to just hedge and buy like a $5,000 node and like plug it in the box out office with wired gigabit or something or at my house with wired gigabit. don't know. So I've been kind of toying with that. That's part of the reason I'm setting up the box out brain. Jeremy Smith (42:47.816) Yeah. Huh. Jess (42:49.079) Wow. Jess (43:12.845) I mean, can you run like anthropic models locally? John (43:17.431) Not entropic, but GPT OSS is there and then you can run. Yes, but the open source models like Kimi, Minimax and some of them are testing at the same as Claude are pretty darn close, like close enough that you won't notice the difference. So that's, mean, and then you can actually run them in Claude code. Like you don't have to run. Jess (43:20.383) Okay, so they have to be open source models. Jeremy Smith (43:38.054) Mm-hmm. Yeah. John (43:39.629) You know, Claude will let you swap out models. So if you download like Olama, you can just do pick a model and do like Olama Claude this and it'll just run that model. The problem is that, you know, you'd want to do that on a server that can actually, you know, they can do it because otherwise you're just using like Quen code or things like that, which I haven't found to actually be great compared to Claude. So Jeremy Smith (43:55.27) Yeah, beefy enough. John (44:04.343) Yeah, but definitely I bet that's one thing I've been kind of think about because I'm like worst case scenario, you bought a 5k server and you had some fun with it. You know, best case scenario, like it doubles in value over the next year and continues and then you can run your local models and the models will just keep getting faster and faster and maybe you can serve your whole team. I don't know. I don't know exactly how intense it is to answer one of these questions, you know, so. Jeremy Smith (44:27.506) It is super weird that we were getting to this point where I know some guys were bragging they were working off of MacBook Airs. Minimal tools, low power devices can

do everything, and now it seems like we're suddenly going back the opposite direction, which is like, boy, maybe I need more power in-house. Maybe I need more equipment. I used to have so much hardware in my house for like... doing computer-y stuff and now like, and then I like, you know, streamline, right? Yeah, no doubt. Yeah, the number of services I have, yeah, yeah, Mm-hmm. Jess (44:59.776) Yeah, and you can't even code offline now. John (45:03.711) It's yin and yang. It's just the pendulum swings, you know? So the pendulum swung to whatever mainframe with clients, and then it swung to have your own hardware, and now it's going back to mainframe with clients. So for me, the last five days, was like, I don't need 64 gigs of RAM on my laptop. If I just set up, I mean, I went and grabbed a Hexner box for like 30 bucks a month and it's got like whatever 16, 32 gigs of RAM in the US and like the US cloud. And I was like, this is fast enough, honestly, like for the stuff I'm doing, that's the box out brain. And so I'm like, what, I mean, I can stall all my ZSH, I can stall everything that I want to do. don't, the blocker before was like editors, but I don't use an editor anymore. So I'm like, that's not a blocker. And so it's just like, Jeremy Smith (45:50.802) Yeah. John (45:56.151) I mean, why not just have like an air and then it's always on, you know, you can always connect to your brain whenever you want and your phone can connect to it and everything can just get there and, and you don't have to swap sessions because like it's just there and you can just see it in both places. Like I feel like that's the short-term future. Like in the next couple of months, someone's going to solve that really well. then I have a Jeremy Smith (46:15.208) Yeah. John (46:17.056) like, you know, conductor, you know, is one that could solve it really well. There's happy to engineering, which is an interesting take, but I don't understand the security model on it or any of that kind of stuff. I haven't looked at it enough. but yeah, I don't know. That's, that's the next thing I've been thinking. Jeremy Smith (46:26.16) Okay. Huh. Jeremy Smith (46:32.872) Can we change gears? sorry. Jess (46:33.036) Can your conductor connect to your brain? John (46:38.55) So. Jess (46:38.944) Or is it only local? John (46:42.188) I think my gut says right now it cannot, but if Conductor does certain things that they're saying they're going to do, I don't know if they're saying it publicly or not, which is why I'm being weird about it, but if they do certain things, then I could see it being more possible, or if they just solve that problem. All I want is I want to be able to get to you have some server somewhere that's always on and always working for me that I can be notified on my phone and my laptop when it's done and prompted to do the next thing. And if necessary, spin up a server and, and tail scale, see it on the port that it's running on, on that machine.

Like that's basically all I want. could, I could glue a bunch of crap together right now, but it would, I'd say it'd be manual. I'd have to tell all the time. I had to tell like open code web to like, you know, spin up a workspace and a work tree and then, you know, to run the server, do this thing. And like, I'd have to do some of that junk right now. I'm sure some people have figured this out. I just, I'm too busy getting stuff done to like, you know, iterate on the process more than I am currently, but that's what I want somebody to solve that hardcore in the next like month. Because I would, I would, I would pay as much as I do for Claude or anything else for that solution right now, you know, Jeremy Smith (48:03.867) Yeah, that also gets it away from, it solves some of the security issues around running in like YOLO mode on your own machine, right? Like, so if you've got it sort of sequestered on his own server, and it's only for those purposes, you've kind of tied off one area where you could get into trouble, I guess. One thing that I, yeah. John (48:26.879) Yeah. That's why I'm doing the, Hetzner bots. Cause I was like, let's just throw it on there and worst case scenario, I'll rip it down. I'll just delete it and create another one. And instead of just manually running, like somebody was like, just install Claude and then let Claude set the whole server up. And I was like, yes, but I'm going to do that from Ansible on my lap, on my laptop, you know? So now I can just like capture on our deploy Ansible to deploy. Jeremy Smith (48:33.403) Yeah. Jeremy Smith (48:46.79) Yeah. John (48:51.447) and like I have vaults for the all the tokens and all that kind of stuff and it works super well like to you know so I like literally it would like deploy and then it'd be like oh the deploy failed and it was in a loop and it would fix the deploy deploy succeeded like you know and then it would go and like SSH and verify that like the processes are running and do all these things I wasn't telling it to it was just figuring it out Jeremy Smith (49:05.521) Nice. Jeremy Smith (49:12.167) Yeah, nice. John (49:14.249) And it was in this perfect loop where it was just like solving the problem. And then eventually it's like, cool. Yeah, it's all working. You're good to go. Like go to this URL. And I was like, that worked really well. So like that's, want that. But like for my dev environment being set up like everywhere and being able to access it on my phone and on the laptop, whether it's web, it does not honestly webs fine. I don't need Mac apps. I don't need other stuff. Like I think if you do it right, it could, it could be okay that way. So, but yeah, anyways, you said you want Jeremy Smith (49:38.853) And what you want from the web view, what you want is to be able to see the app running. Is there something else you want to be able to see the chat session as well or just the app? Yeah, okay. Chat, Yeah, okay. All right. John (49:47.346) Yeah,

that's what I want. I want conductor in the web is what I want. So the ability to start processes, stop them, like all that kind of stuff and then be able to chat with it and prompt it on the next thing. like I really feel like conductor in the web or. Jess (49:51.638) Yeah. John (50:04.232) If it's Mac and iOS, doesn't really matter, but I'm okay with web. I'm a web person, so that's probably why. But I think, I think that's what I would really like. And it doesn't even need some of the stuff they do currently, I feel like. But yeah, just, just the ability to like have my laptop just sitting up here and be like on my phone and just like keep spinning up new things. Like, I mean, I have a to-do list that's huge. Like I could just literally just be like copy paste into new workspaces, all these items and just have a thousand Jeremy Smith (50:25.509) Yeah. John (50:34.376) not a thousand, but hundreds of workspaces just working on them. And then I could just go back and like verify them at some point. And you know, with playwright, you can automatically browser click and take screenshots. And so then it's just like on a lot of this stuff, I could literally just look at images at the end and be like, Merge, this is good. Like I've looked at the code, I've looked at the images, I don't even need to spin the server up. But Jeremy Smith (50:36.135) Mm-hmm. Jeremy Smith (50:59.239) I feel like Avdi Grim called this, I can't even remember what he called it, was like mom driven development or something like this where basically like he's doing all kinds of things taking care of his kids, but he has an idea, like he's making dinner and he has an idea for something he wants to get done. It's being able to spin something up and know that it's gonna take some time, but then it comes back and you've just got these like little blocks of time. I have another friend that's like, yeah, I finally, John (51:15.507) Yes. Jeremy Smith (51:26.651) got the service so I can SSH from my phone into my computer at home. Cause if I'm in car line, you know, like I can check on where things are because if you're moving into more of a management mode and you've got something else that can take over a lot of these things for you, you really only need these smaller touch points. And it doesn't make sense to that. It all has to fit in a nine to five work day, which we could talk about also the, the way that this bleeds out into us working all the time. John (51:31.167) Yeah. Yeah. Yeah. John (51:52.757) I'm not worried about the health and safety right now. Jeremy Smith (51:54.92) I think there are different models of work and some work for some people and some work don't. I grew up in farm country and John, know you did too. And all the farmers I knew like lived, you live in the center of your farm and the work is never done. know, like, you know, like it's you're going out to do chores very early in the morning, coming back in for breakfast. You're out late at night.

John (52:07.753) Yeah. Jeremy Smith (52:23.473) You're doing stuff on Sunday. It's all around you all the time. And it's just woven into your entire life. I'm comfortable with that. I know for many people, that's not a, that's not a way they want to work. and that's totally fine. But for me, I like, I like being able to say, let me drop in and do this for a second. I might be, you know, pushing PRS at 10 PM. You know, it's just like how I work. I'd rather not have my time. Like, you know, eight to five, like don't talk, don't John (52:29.598) It is. Yeah, sure. Jeremy Smith (52:51.503) None of my personal life can invade in that time. Like I'd rather have it mixed up. but I know that doesn't work for everybody, but this model of being able to say like, I'm on the go. I just want to like check in on this thing and like nudge it for the next little thing that it can do. Because if I wait, you know, then I, you know, then I have a bunch of things to plow through much later. But if, if I can just come in and fix this one issue, then it gets the work done faster in the calendar time sense. John (52:56.99) Same. Jess (53:21.033) Or if it just gets hung up on a permission, you're like, ugh. Jeremy Smith (53:23.119) Yeah, right. John (53:23.506) Yeah. Well, I honestly think that it's a more healthy way. Like, I mean, I've specifically told everyone on my team, if you see me doing crazy things all the time, just ignore it. like either work, you will know we're cool. Like you're fine. But this is just my workflow because but I think how it makes it more help, you know, a it makes it unhealthy for me because I'm carrying around this laptop all the time. So maybe I need to get a MacBook Air and have a beefy server somewhere. So that that's one thing. But but if we solve that phone problem, then that problem goes away because it could be phone or iPad or whatever. It doesn't matter. It doesn't have to be like this heavy device with a screen open that's more, you know, blocking to like other things and stuff like that. But I'm the same way I grew up in the farm. Jeremy Smith (53:38.726) Yeah. Jeremy Smith (53:46.779) Yeah. Jeremy Smith (53:56.039) Mm-hmm. John (54:05.488) the same way like it's a lifestyle and like, in the same way that like, I don't know, I'm curious if this was the same way for you. But like, you know, was there always work to get done? Yes. But if my dad ran into like another farmer, did they talk for an hour on the side of the road? Yes, they did. And like, I think, you know, the point is, you should be able to do the thing whenever you want or need to. Jeremy Smith (54:18.711) Right. Right. Jeremy Smith (54:26.854) Mm-hmm. John (54:27.316) But you should also not feel like you have to do it 24 7, you know, like I have a lot of long hour long conversations where I'm trying to convince someone why conductors better than whatever they're doing. And like, you know, eventually they all give in and they agree with me and it's fine. But Jeremy Smith (54:36.801) Yeah John (54:42.772) You know, I couldn't do that if I was like, well, it's night between the hours of, you know, my, my whatever eight 30 and three o'clock. So this is all I can do. I have to be heads down on my own things. Instead I can say, no, I can have this conversation because they're here right now and meet space and I can enjoy this. And then later on, when I'm watching outlander with my wife, I can type a prompt in and work gets done. And that that's I've to me that's fine as long as it's not like I'm not always doing that. Now I've definitely went a little crazy the last month or two so I can accept that and you know from that standpoint but I think if we fix the we meaning the world and Claude and whatever the this iPhone and other like other ways of nudging things along that it will be it will feel more healthy and normal for me personally so. Jess (55:31.731) It's different when it's optional. Like you, you want to do this. This is on your mind. So you, you're, you're logging in and you know, pushing the button. It's different than like somebody calling you up and demanding something like you got to do this. We need this get to work and John (55:34.182) Exactly. I do. John (55:44.957) Yes. Jeremy Smith (55:48.327) there's still an impact on your own relationships, and how that's, what that's communicating, whether it was like you taking a work call or, you know, customer call in the evenings, as opposed to now, like checking in with this agent on this piece of work that's still, you know, like that still communicates to other people, like your attention. And so I think that's, that's always a concern. It's, that's not, not a new thing. It may be new in this new way, or maybe have a temptation that's new here, but. John (56:00.627) Yeah. Jeremy Smith (56:17.734) that concept itself is not new and has to be taken into consideration, I think. Jess (56:22.731) I just like how we just, so the, podcast is like the same thing as you running into another farmer on the side of the road and a tractor and talking for an hour. Isn't that cool? Jeremy Smith (56:30.766) Yeah. Yeah. We are in normal business hours right now. Yeah. John (56:38.258) Very much so. But again, like, I think this is aw- I think it's awesome. Like, I think it's great that, Jeremy Smith (56:39.675) Yeah. John (56:45.072) you know, you can just sit down or I can just sit down and join you guys for like an hour and we can talk and we can go over this stuff. I mean, I think that that makes things better, like in the end, like, but yeah, I mean, there's definitely like, there's some unhealthy bits to this and stuff. And we haven't even talked about, you know, like all the long term, I'm like, where's this going to be in six months or 12 months? Like, well, will my job exists? Like, well, probably not. It'll be in a different way, you know, but so am I accepting that and going into a different way? Or am not, you know, I've

had this conversation with, Philip who works with us. He's like, I was like, Hey, how much are using Claude to build our, you know, to build this? And he's like, well, not too much. And I'm like, why? Well, like, I like to write CSS. And I'm like, I hear you. What if you, you know, like, I, I can understand that I get it. And I used to think I was a craftsman, a craftsperson, whatever. But now I'm like, no, I just really like solving problems for people. Jeremy Smith (57:30.746) Yeah. Mm-hmm. John (57:41.265) That's what I like doing. And now I can just do more of it at a higher level. And so, yeah, I mean, but I've also been on this trajectory for like the last, let's say, five years trying to learn how to manage a little more and some of those kinds of things. So maybe that makes it feel more natural or better or whatever. don't know. But. Jeremy Smith (57:57.221) I think there is another thing here which maybe, I don't know, it just occurred to me that there is a difference between, if my work changes from I am writing code, I am producing features to I am telling Claude to produce features for me, there's still now a difference between owning the product and then working for somebody, for an owner. where it feels better if you are, I think it feels better if you are the owner doing this than if you're not. I could be wrong about this, but I think it probably translates into like, could have some better feelings about my work as a non-owner when I was the one creating all the output, if that makes sense. And now, John (58:32.006) Mm. John (58:49.842) Hmm. Jess (58:50.9) or you have more leverage. Jeremy Smith (58:53.046) It's not, but it's, think it's not just leverage. It's also what I'm shaping. If I'm shaping the code for this feature, that's like my imprint on it. But if I'm not an owner, like if that goes away, then I wanna be an owner because then the thing I'm shaping is the entire business. Does that make sense? So I feel like I've, I think I've already noticed this, that there's like a, it seems like more pressure toward, John (59:14.876) Yeah. Yeah. Jeremy Smith (59:22.254) more people being owners. That would be my, I don't know, that just feels like for developers, the place to go is more trying to build their own products, becoming owners. John (59:35.953) Yeah, think the problem is most people just don't want to do that. Like they just they just want to show up and do their job and go home. I think the you know on the bell curve that's probably a lot of the people 70 80 percent or more. So I don't know what that's going to mean for that. What I actually thought you were saying is that you enjoy it less when you're shipping code for customers when using Claude. Like is that is that what you're saying. Like does it feel like it does. OK. Jeremy Smith (59:38.629) Maybe not. Jeremy Smith (01:00:00.611) Yeah, yeah, yeah, there's, mean, I'm doing it because I think it's the best way to use their, you know, to use the resources to, you know, like to deliver value for them. But I, there's maybe some satisfaction and be able to get a slightly more work done or be able to work when I'm tired, like where my brain is just like, man, I don't, I can't even think anymore and be able to still be able to get things done. But John (01:00:06.01) Yeah, of course. John (01:00:10.105) Yeah. John (01:00:24.335) Yeah. Jeremy Smith (01:00:29.23) it feels less, I feel less work satisfaction, I guess. Yeah. Because it's. John (01:00:35.567) Hmm. Do you do you charge based on you just charge based on hours? Yeah. Because I've been thinking about that a lot. There was a side of me that was like, maybe I should be an AI consultant. Jeremy Smith (01:00:40.334) hours. Mm-hmm. Yeah. Which I like doing it that way. Yeah. Jeremy Smith (01:00:48.261) Mm-hmm. John (01:00:49.775) I mean, not that I would stop doing like my other stuff or anything like that, but there was a side of me like I can literally build someone an app in a weekend in like three, four hours. That's totally worth five K to them, you know? And so like there's a side of me. It's like, maybe I should try doing that. Like, and because I have enough stuff I own, so like I don't, but that would kind of help me on like zero to one of like get tuning in on like how to build new apps from scratch more quickly. cause I think the better you get at that, the more you could solve problems for people. Jeremy Smith (01:00:56.773) Mm-hmm. Jeremy Smith (01:01:00.719) Yeah, right. Jeremy Smith (01:01:07.845) Mm-hmm. Jeremy Smith (01:01:11.888) Yeah. Jeremy Smith (01:01:15.483) Yeah. John (01:01:19.698) people. So I've definitely thought about that there and now like what I my wife was like you should definitely do that and like teach other people how to use AI and all this kind of stuff because you won't shut up about it. But like there's this side of me that's kind of thinking like that's my backup plan. It's like well if all my businesses go to zero then I'll that's what I'll do. I'll I'll you know I'll just do that. I'll build custom just bespoke you know solutions for people or come into their company and show them how they can use it or things like I'll do that if everything else doesn't work because what I like is building. just like building. Jess (01:01:29.834) You Jeremy Smith (01:01:33.732) Yeah. Yeah. John (01:01:51.716) I don't care if it's building for somebody else or myself. I just like making software that solves a problem that people get excited about. So yeah, don't I mean I do like a minor tiny bit of consulting and it doesn't barely count because it's on speaker deck, which I originally created. So I just do it for a friend like as and so but I'm like I totally love it like and I still build based on ours too. guess on that when I do but Jeremy Smith (01:02:05.679) Mm-hmm. Yeah. John (01:02:18.062) Yeah, I don't know. I feel like if you were really going to go for leverage, you could totally be

like, well, like this feature would normally take me 20 hours. I can do it in two. I'm not going to charge 20 hours, but I am going to charge 10 because I learned how to do it with AI to make it more efficient. Like if you had learned how to you learned how to code faster. And I assume at some point you raised your rates and you got more experience. You get, can dive into it quicker. And you can, you kept raising your rates. This is just another rate raise. Like almost is what I would maybe think of it, but Jeremy Smith (01:02:44.941) Yeah, I think I would probably, I would probably figure out a new service to offer that was like this. Like if it's the zero to one kind of like app and a weekend kind of thing, like I'd probably do that different than I wouldn't want to change the sort of the standard relationship I have with clients currently. Cause I still like the idea of the way what I'm selling is partly delivery of code, but partly it's like, John (01:02:54.297) That makes sense. Jeremy Smith (01:03:13.345) you need to get out of a lot of the heavy lifting in the code base yourself. You need to hand it off to an experienced developer to ship features and get maintenance worked on, all of that, and to really kind of have the understanding of the whole system in somebody else's brain beyond you. And so I think there's more to it than that. that's, I don't want to mess with that service. So I might try something else. I could, I could imagine doing that. I think that's, that's probably a good idea. Like the, definitely makes the prototyping or the, MVPs much faster or things that are much, that would never have been done. Like makes them suddenly the economics kind of work out on that, better. John (01:04:01.997) It doesn't even have to be zero to one though. Like I'm just thinking if I were you, like just truth talk, I think if I were you, I would be billing more than the actual times that I'm putting in because you're creating way more value. if it, you know, pre AI, the amount of time you would have to spend to build whatever some system was going to be a hundred hours. And now you can do it in 10 and the end result is basically the same. Like I it's only fair that you receive some you don't have to get a hundred hours, but you should get some, mean, 20, 50, somewhere in that range feels like an acceptable amount for you to increase to be like, look, you're getting more things. Even if, know, you could just do a personal like, check of like how much more productive am I actually like how much would this have taken before you know but I just feel like when there's efficiencies that are gained that you're like a master of or you control like that's a benefit to everyone who's using you so if you don't charge more for that then you're basically the one that's losing on it in I don't know that is most simple way to look at it. Jeremy Smith (01:05:05.219) Yeah. I think I haven't seen it long enough to know how, how much like to I'm feeling

faster, but I also believe that there's a possibility that people feel faster when they're actually not. So that's one thing I'm still not entirely convinced of that sometimes things feel faster, but actually you were just, you know, you were just spinning your wheels. Like you were actually like, I don't know if you've seen this concept of like, John (01:05:13.679) Mm. Jeremy Smith (01:05:34.98) work in progress that like, if you really want to increase the throughput, you'd reduce work in progress because it reduces contention as you're moving back and forth between, you know, very quickly between lots of different things. And so I found that when I reduce work in progress with my clients, and for years, like this has been the pattern for me, that I actually ship things faster. Now I feel like I've got more work in progress and it feels like you're just busy because you're like skipping across all these different things back and forth but is it actually faster and that's like I think hard to measure because it's not just yeah yeah I mean I know you're convinced I know you are yeah okay John (01:06:12.673) Yes. John (01:06:17.921) I like I have a script though I have a devlog script that I that I had Claude make and it shows you know like it shows my basically my pull requests and my to do is checked off from to do list and like all that kind of stuff. And it kind of, so I can say dev log seven and it's like, here's the last seven days. And it's like, here's all the issues created. Here's the pull requests that I opened and didn't merge. Here's the pull requests that were opened and I did merge. And here's the ones that I did both up in the same seven day period. And like, it does all that stuff and it is significantly more packed and everyone around me that I'm working with who is not a developer who's not using AI like this. like, my gosh, they're like, they're literally blown away by like, you already got that done, like, and so I'm like, it definitely is how I'm not saying that that is happening for everyone or not, because that's not for me to decide. But like, I based on the checks I've done on myself, I am definitely getting a lot more done than I was three months ago, like to Jeremy Smith (01:07:02.179) Yeah. Jeremy Smith (01:07:20.613) Are you working more? These are other things I'm wondering. Because you're so excited and you're working with your energy, is the current, what you've been doing in past few months sustainable? Can you actually keep the space? Are there any externalized costs that are happening psychologically or with your own energy capacity that do you burn out? I feel like so many developers are so frantic right now because John (01:07:49.166) Mm-hmm. Jeremy Smith (01:07:49.178) We're trying to find out, can we be faster with this? We're also trying to figure out, we're also excited. We're also trying to like make sure we have jobs. So we're all working more. I think we're working more and we're working harder and faster. And then probably, you know, everyone else is really excited about that because, Hey, look at all these productivity gains, right? But then I wonder if there's, I'm talking to another friend, like maybe there's a crash in six months where all the developers who have been working so hard, both excited and nervous, scared about the future. trying to preserve their jobs, trying to figure out where we're going. We just all burn out in six months because we've been working our tails off and at a pace that is not actually sustainable for our biology. John (01:08:19.532) Yeah. John (01:08:24.417) Mm-hmm. John (01:08:32.348) Yeah, that's a great question. I definitely think that's that's probably gonna happen for like some number of people and stuff like that. I don't feel like anything I'm doing is unsustainable. were you saying, Jess? Jess (01:08:37.769) percentage. I'll just add some percentage, yeah. Jeremy Smith (01:08:43.171) Yeah. I can tell that you're excited. You're happy. You're excited. I hear you're hedging in some ways maybe, or you have plans and things that make you concerned, but mostly you're following your joy and excitement and curiosity. And that seems like helpful. I think a lot of people aren't necessarily there. They're not totally following all of that. Yeah. John (01:08:43.531) Yeah, yeah. But like, like John (01:08:58.174) Yes. John (01:09:03.287) Well, here's the thing. just I'm just looking at it as I feel like there's an arbitrage for like the next six to 12 months. I don't know how long it lasts, but there's an arbitrage where you can actually get more done. You can actually do things you couldn't do before. And because that there's more work in progress, but that work in progress doesn't require your constant attention. So you're, you are actually creating more artifacts that are usable. Now there is still a constraint on how many of those artifacts make it to production and actually create value. But I think there's also value in exploring those other things, even if they don't make it, you know, that's fine too. But, and I'm definitely, I agree with you. I'm definitely excited. I'm drinking the Kool-Aid. So, you know, nobody has to listen to me. Jeremy Smith (01:09:34.947) Yeah. Right. John (01:09:49.902) But I what I keep thinking is there's like six months or 12 months. You guys have seen the graph of like it's like all the checkbox, just the boxes and like, you know, it's like there's a checkbox for every three. Well, no, it's for every three point two million people. There's a checkbox in the in the world. So eight point one billion people. There's a checkbox for every three point two billion. And the gray grayed out is like people have never used AI. And then green is like people who have like talk with a chatbot. Yellow is like people paying and red is like actively using it to build. Jess (01:09:59.672) yeah, the weak. John (01:10:20.042) stuff and and like drank the Kool-Aid

appropriate that's red because it's Kool-Aid and that so like the red block is one out of like like thousands you know or whatever it is and then the yellow is like six or seven the greens maybe like two to three rows and then there's like 80 rows or something of just like people who have never even used this before and I'm like what we're seeing right now is just from the nerds you know like what happens when all the non nerds Jeremy Smith (01:10:29.73) Yeah. Jeremy Smith (01:10:46.083) Yeah. John (01:10:49.172) or like the half nerds and all that start getting into it. Like I you know when I talk about the stuff everyone's like you're crazy but as soon as I show them they're like my goodness I get it. You even I'm talking parents you know like you know that like my mom used it to like they have a. Jeremy Smith (01:11:00.056) Yeah. Yeah. John (01:11:11.624) And she was like, that's awesome. And that's what they gave to the contractor. And that's what the contractor did. And they look identical. And they were like, well, what about the barn? What could the barn look like? That's right beside it. And it was like this. And now it looks like that too. And I'm like, and that's where she was like, I used AI. That was amazing. You know, like, and I'm like, yes, you know, my father in law is all like, no anti AI on some stuff. And I was like, I was like, Brady, no, I'm like, you're not, Jeremy Smith (01:11:13.988) Nice. Jeremy Smith (01:11:24.142) Let's go. John (01:11:34.317) I was like, what do you want to know about anything? You know, and he's like, I was curious about like the gear ratio on my truck. And I was like, okay, just ask. I asked AI took a picture of a thing. AI told him the exact gear ratio. And here's the up and the down. This is why his gas mileage isn't as good. And like all that stuff. And he was like, oh, you know, yeah, every gets I kind of like, okay, you know, reaction, we all remember that. And I'm like, that's that kind of stuff. I feel like Jeremy Smith (01:11:36.099) Yeah. Jeremy Smith (01:11:52.965) Yeah, yeah, yeah Jess (01:11:55.177) You John (01:12:02.877) again, it's just the nerds. Like we have this arbitrage right now. Arbitrage is a point between when you know something and other people don't where you have an advantage, you know, and I'm not trying to keep the advantage to myself. I'm talking to everybody about it, especially my friends who are not programmers or whatever. I'm trying to say like, like you need to start using it now. I just had lunch with the guy who's not a programmer at all. And I was like, you need to be in Claude doing stuff right now. Like Jeremy Smith (01:12:15.384) Mm-hmm. Mm-hmm. John (01:12:29.031) learn things, you know, like on your own time, like it doesn't have to be it can be at work, but either way will benefit. I know it will benefit you. I don't know what it looks like. I mean, I'm terrified that like I'm just trying not to think about that because I don't have any control over what the world looks like in six months or 12 months. Like I have no idea what it will be. I don't know what jobs will go away, what jobs will sprout up every time there's you know, the thing that gives me some solace is every time there's a technological improvement. Typically, everyone as a whole is better off. But some people get left behind and it's horrible. So I don't know. I don't know about that. don't know who it's going to be. I don't know how that's going to work out. I don't want it to be me. So I'm trying not to have it be me. And I don't want it to be anybody that's around me that I'm friends with that or that I know. So I'm trying to get on them, you know. But that's all I can do. I don't I if I I'm just not like I would say I'm not. Jeremy Smith (01:13:19.534) Yeah. Yeah. John (01:13:26.93) I'm an optimist generally like I give you if people talk to me they'd be like all you do is point out everything that's wrong. Well, yes, I do. But I also believe everything's fixable too, you know, and so that's that's what I'm trying to focus on myself is just like Focus on the positives right now. I've got this cool tool that helps me get a lot more done. Verifiably, I can I can show you like if I just sat down and went through everything I've built in last two months, have I worked more? Let's say I've worked if one was my normal output. I've worked one point three to one point five. Let's just say that I did that. That's possible. But I have shipped 10. So like I could just cut back to one. But let's look at every labor increase in the whatever the past thousand years. It's always by a technological Jeremy Smith (01:13:55.715) Mm-hmm. John (01:14:08.42) advancement and the labor force never works less. They always work more. I mean that is kind of always what happens. So guess we have to keep that in mind. But now I'm just means I'm working like a more normal work week. Before I was working like only eight thirty to let's say two thirty or three which again I know I'm fortunate that I can do that but. Jeremy Smith (01:14:23.236) Mm-hmm. John (01:14:26.838) now, okay. I here and there off and on, you know, five minutes here and there throughout the evening and stuff like that. it's not like I'm, I don't feel drained. I don't feel like I'm hurting myself. I don't feel exhausted. I've heard people that are saying that and that have taken the weekend off and like that came back energized. I haven't felt that yet, but I'm sure that I will at some point. I think my moods are just more. the sine wave just in general, like every six weeks, I just feel like everything's going to implode and then I wait two days and then I'm ready to take over the world again. It just happens. I just recognize that's my, that's my wave. And so I just, when it happens, I tell myself this, this will not last and then pull myself out of it and move on. So I'll stop ranting now, but Jeremy Smith (01:15:06.178) Yeah, it'll pass. Yeah. Jeremy Smith

(01:15:12.012) I think, yeah, resonates a lot of what you're saying resonates. I don't know what to think about. I've thought about this arbitrage thing as well. it makes sense. And I think leads to some of the franticness that developers who recognize the sign of this, what it means could mean like in some ways a gold rush, but you don't know where the gold is. And there's a big part of me that's also like, you know what? I don't need a gold rush. Like I don't need, that's not who I am. I wouldn't have moved. to California to pan gold. I'd rather just do work that I like for people I like and try to get, live a good life. That's for me, that's what I'd like to do at the same time. I don't, I want to sort of, you know, protect the, the business that I've developed and, and, you know, be able to retain the skills and, do as much of the kind of work that I've enjoyed doing as possible in the future. So I want to keep up. But it feels like everyone's trying to figure out how to like. John (01:16:09.607) Mm-hmm. Jeremy Smith (01:16:18.382) build something uniquely possible in this time, right period, this period right now that, because we've seen it before, we've seen it so many, if you've been in tech long enough, you know like, man, could have thought of that. Why didn't I think of that that last time? I could have done that. That's happened to me a number of times where you just know, you're almost like, I don't know, just like shaking, like what is it? What is it? What's the thing? know, going to bed, like what is the thing that I do right now that would, that I could, if I built it, man, that would be, you know, it would be the thing everyone would be in three months and be like, my gosh, of course we needed that. I don't know, that's a weird situation, I guess. John (01:16:56.956) The thing- John (01:17:00.602) It's very weird situation. I feel the same reverberations of like, you know, terror slash excitement on those things. The thing I've been thinking and I don't know, I'll again, I'll just say everything I'm thinking. And if I was right and this benefits other people more than me, fantastic. You know, like the thing that I keep thinking is like, well, what will the AI world need more of? So like, try to think from the beginning of like, okay, let's imagine a world where more people are shipping more code. What? increases if if that is true and I think the thing that increases is monitoring and debugging and Like developer like good developer tools and experience or where developer might be random person, you know, whoever like more AI friendly things for actually when something's not going well like So that's, that's like, that's what I've been thinking. So that's why I've been like, I realized the other day, was like, well, I can't you in the old era, I could not, all about Taylor Swift. So in the old era, you know, the non AI era, it's like, okay, what Jeremy Smith (01:18:02.04) hehe John (01:18:06.16) I can only have a certain number of apps. Like I can't actually have more. Like there's only so many I can maintain securely and all that kind of stuff and resilience and performance and, that kind of thing in the new era. That's not true anymore. Cause like we've got really easy ways to host things we've got. It's relatively easy to like upgrade, you know, rails and deploy and all that kind of stuff. and so like a lot of things that weren't, you know, weren't true are true now. Like now you can't. you can invert it and you can say, actually, I can have more apps. And actually, it benefits me to have more apps because so like the simple examples, like I have flipper, it's a dev tool. And it like, I mean, it's been growing really well, especially the last two, three months, like it's almost all of the doubling that happened last year was in the last two or three months. And I'm like, that's interesting to me. And I think that's because more people are shipping stuff, and they want a way to do it safely. Jeremy Smith (01:19:01.815) Yeah. Yeah. John (01:19:01.999) And that's what it helps with. And so I'm like, well, what's another thing you want? You want APM stuff like, and so that's why I've been working on one of those. just deployed it. one of my apps to use it for the first time. So in production, it's starting to use it and record data and stuff like that. Jeremy Smith (01:19:07.235) Mm-hmm. Jeremy Smith (01:19:14.391) Yeah, that's cool. Yeah. John (01:19:19.065) this morning and I haven't clicked to see if it actually worked yet. Cause I hopped on this call with you guys, but, but like, I, I'm, makes sense to me. I'm like, people are going to want more APM and they're going to want it more cheaply as well. Like they're not going to want to spend a fortune to pull in everything. They're going to want very specific opinionated things because they don't know what they want and they don't know what's important. And so I feel like what they want is they want somebody like me who has 20 years of experience and worked at GitHub and all that kind of stuff to be like, this is what matters. Jeremy Smith (01:19:30.754) Yeah. Jeremy Smith (01:19:35.085) Yeah. Jeremy Smith (01:19:48.547) Mm-hmm. John (01:19:48.84) And this is how it's exposed to your AI. And so because it's exposed in this way, you can just be like, sight slow, what's going on? You know, like that's, that's the world that I want to live in. think people in the near future are gonna need to live in or want to live in. that means that's going to happen more, you know, like GitHub scientists where you can run two flows of code at the same time, compare the performance and compare the results and return, you know, the original flow. So you know that you're returning the right thing. I think that's what I'm going to, I'm going to build after that. Cause I'm like, I think that's also important, you know, Jeremy Smith (01:20:10.305) Mm-hmm. Yeah. Jeremy Smith (01:20:20.353) Hmm. Yeah. John

(01:20:23.896) So that's in my head. That's what I'm thinking is like, actually people are to want more developer tools and more like less UI. I'm build the UI because it's free. Who cares? But they're going to want less UI. I think they're going to want more integration of like what excels in the AI era flat files. So they're going to like in what I'm planning on doing is cheaply storing a bunch of stuff in cloud storage and making it really easy for the AI to pull down what it needs to investigate. Jeremy Smith (01:20:25.923) That makes sense. Jeremy Smith (01:20:53.325) Mm-hmm. John (01:20:53.695) I mean, think that makes sense in this era. So if somebody else builds it first and or does a better job, that's fantastic. But I don't think any of the incumbents have any incentive for you to send less data because that's what they charge on. So in the AI era, you invert it and you say, well, if it's really expensive right now and you send everything, how can we send less and make it less expensive and more directly integrate it with, you know, Claude or tools like that? Jeremy Smith (01:21:05.795) Right. John (01:21:19.918) I think that's that that wins in the next iteration. I don't know for how long, but in the next one. And then if you have two developer tools, flipper and caboose, whatever, you can send them back and forth. so you can cross sell and you know, then you add whatever, like a way to not only just run this code code or the other code, but to run both at the same time and collect the samples and, help with all that kind of process. I think that's another thing that just fits in. It's one step away in the ready. Right. It was at ready fire aim that book, you know, it's always about like one step away from your current market that you're Jeremy Smith (01:21:29.688) Yeah. Jeremy Smith (01:21:43.959) That's nice. Yeah. Jess (01:21:46.437) Yep. John (01:21:49.832) marketing to and we're doing the same thing on box out same approach is like what are the schedule what are the other things that these high school a DS and and you know sports information directors in college what did they need what tools they need and I think that's probably you know what you have to do is just look at those things if you're if you're doing what I'm doing if you're consulting I think you raise your prices or you keep your prices the same and you are basically doubling your hours like and I don't think that that's a as long as you're actually delivering more value it's not dishonest in my opinion Jeremy Smith (01:21:50.135) Yeah. John (01:22:19.832) you're delivering value or maybe you have a conversation with your client and say hey do you feel like I've been shipping more because you know etc etc I don't know what that means or more clients Jeremy Smith (01:22:29.315) It may be making a great, yeah, think it may be agreements around like, if you want me to use agents, here's, if you're like, if you, or if you don't want me to use agents, here's the price. I don't know. Like it could be something like that. I mean, something like that. Yeah. Um, yeah, it becomes a new, yeah, certainly a, a new discussion, a sort of agreement, although up to this point, I haven't had those. think everyone's sort of just assumed that, but yeah. John (01:22:41.86) Yeah, it's like credit card, no credit card, like the cash discount, whatever. Jess (01:22:44.346) This is one price. Jess (01:22:57.552) Figuring it out. Jeremy Smith (01:22:59.671) This also brings back to mind the idea of farming for me, which is like, like years back, if you have a family farm, there's only so large it can be if before tractors and combines. And now like you have tractors and combines, it forces farms to get bigger. You now have a family that has much more surface area to cover to run a single business. Some ways that really sucks. Like it's cool in one way and other ways it kind of sucks. John (01:23:17.273) Yes. Jeremy Smith (01:23:29.613) But you have to like, mean, I knew guys, you know, in high school that they would take days off school to go be on the combine or, you know, like, cause it was, if it was just their dad, doing it, it wasn't enough people. So you have all the boys, you know, like out doing stuff, taking days. Yeah. Right. and in some ways it's kind of cool. I like that. But in other ways it felt, it feels like, man, you gotta like, you gotta own so much land. You gotta have so much. John (01:23:46.863) somebody to run the cart, somebody to run the combine, somebody to run the semi, like, yeah. Jeremy Smith (01:23:59.172) And I mean a lot of farmers are in debt too just like because you have these huge machines I don't know if it's a good analogy, but part of that doesn't feel good to me. Yeah. Yeah, I don't know John (01:24:11.269) Yeah, I hadn't even thought of it that way. That's a really great point. My dad was a farmer and he also I think he realized like it's kind of fascinating because I have not thought about this but I feel like what he did with farming is what I'm starting to do with software in the same way actually because yeah, because he was like, well, you can only make so much money at this size. So the best thing is to get bigger. So and the best way to get bigger is typically to rent because owning is pretty difficult. Jeremy Smith (01:24:32.067) Yeah. Yep. Jeremy Smith (01:24:38.561) Mm-hmm. Okay. John (01:24:39.426) And so that's what he did. And then it's like at a certain point, you know, you get economies of scale, you only need so many combines, and then you get more people and more whatever and, then eventually, like, you know, I mean, he was one of the largest farmers in Michigan. And I think, you know, once you get that economy of scale, it's like, Jeremy Smith (01:24:44.152) Mm-hmm. John (01:24:57.86) You know, it's kind of like Warren Buffett made all his money in his last 10 years, 20 years, whatever it was, you know, it's the same kind of thing. Then you can

actually have profits. You can get negotiate better interest rates. You can buy in bulk for everything more often. And so all these things get a lot better for you. And it definitely is harder to be a really small farmer. You almost have to just do it on the side and have like a day job at this point. Jeremy Smith (01:25:16.771) Yeah. That's I've met. Yeah. Do I did a farm school program 10 years ago in Asheville and, all the people I knew that were running farms that we like went and visited their sites and stuff like that. They were doing, they, had another job. They had a day job at that scale. They had it. If you wanted to do it, that was like, you either had another, you know, like your spouse was, you know, the sole, you know, the primary, income earner or you had a day job. Yeah. So it kind of like forces one or the other. And I don't know if that's the situation we're in. I'm kind of scared it is. It's not necessarily bad to have more surface area. I think what you're saying is like you're speaking to a lot of the tools that would be needed to cover more surface area for any single developer or product owner. But it really changes what it looks like to be doing the work. John (01:26:11.939) It does. Yeah. Again, I totally agree. I see what you're saying. It also terrifies me. But if I just sit there and be terrified, like it's not what's that. That's not going to end well. So the best thing I can think is to learn how to take advantage of it, to do something, to help myself and the people around me. I don't I don't really know any other way. And like even my kids, mean, You know, I mean, my wife's relatively anti AI. I she's probably a little less so since I've had Claude help with some things, you know, in nice ways. But but like, Jeremy Smith (01:26:43.478) Mm? John (01:26:49.089) you know, we talk about it with our kids, like how, what do we show? What do we teach them? What do we show them? All that kind of stuff. And I'm just like, I, I'm kind of of the belief, like they should understand what it is, how to use it and all that kind of stuff. like don't be afraid of it, taking your job, figure out how to leverage it to do something cool and new. and again, I don't know what that will be or how that will work or any of that kind of stuff, but I feel like that's that mindset at least seems positive and forward thinking rather than Jeremy Smith (01:27:04.929) Mm-hmm. John (01:27:15.895) you know, woe is me. Like I was dealt, you know, blows or bad things or definitely not. I mean, it's not a great time to like go into the job market, but it wasn't great in 2004 or 2000 either. Like when when I started working, I mean, it it's tough. I mean, you just got to adapt. Things keep changing. So, yeah, I don't know. It's yeah. Go ahead. Jeremy Smith (01:27:17.046) Yeah. Jeremy Smith (01:27:25.76) Mm-hmm. Jeremy Smith (01:27:35.489) Yeah, I agree. It's like, it's like, we're all playing a hand that we didn't deal ourselves. and there's

no arguing what, you know, like this is here, at this point. so it's like figuring out how to, how to work with it. we don't have to, I don't know. I think it's helpful to look at paths and look at analogies to try to get a sense of what that might mean, but we don't also need to make, you know, gloom and doom predictions about, you know, like, well, it's inevitably going to be. John (01:27:50.029) Yeah. John (01:27:56.866) Yes. Jeremy Smith (01:28:04.886) this, it's going to mean like, you know, terrible situations for all kinds of people. but at the same time we, you know, should be doing everything we can to, you know, take care of ourselves, take care of other people and, you know, try to provide a positive vision, I think for the future. So, yeah, I, I appreciate that you're kind of, you're threading that needle and, bringing a lot of positivity as well as the excitement for, yeah, optimism for it. Yeah. Jess (01:28:26.839) optimism. Yeah, I think developers are very resilient. You know, we've, we've seen a lot of change over the last few years. You know, I mean, go back to the, the personal PC coming out and then the internet. And then, um, you know, we started seeing, uh, the emergence of SAS and then the mobile, uh, market and. You know, now AI and we just keep adapting. We keep learning. And I think that's what developers mostly deep down enjoy the most is learning and figuring things out. So. Jeremy Smith (01:28:34.37) Mm-hmm. John (01:28:34.957) Yeah John (01:28:42.68) Yeah. Jeremy Smith (01:28:47.319) Mm-hmm. John (01:28:52.951) Yeah. John (01:28:56.546) Yes. Jess (01:28:58.777) I would just say embrace that and just keep learning and keep figuring things out and being valuable. You know, just help people figure out how to help people solve problems and just ride the wave where it's going to take you. Jeremy Smith (01:29:06.498) Mm-hmm. Yeah. Jeremy Smith (01:29:13.078) That's actually one thing I keep saying to myself, is like, there's no, there's no limit to human need. just like we're like humans are kind of endless in the amount of needs that we have. there've been so many times you introduce a new technology and everyone's like, well, we'll all be just sitting around now. Like, no, it just like, it just leveled up the number of needs that came on top of that thing. All these time-saving labor saving devices led to new needs. so. new needs need to be addressed by people who can meet them. And, and usually there's value exchange there. So I don't know. That's, that's something I always hold onto. John (01:29:51.265) there's there's two articles out I'm sure you guys have seen them but I'll try and send them afterwards but like the one article was like that the one guy on X that went completely viral the one day and he basically like was like you need to right now is the time you need to tell everyone you know to get on the train or it's really gonna be really bad for them and I was I read it at like 1130 at night and I was like this was not nighttime reading I'm like I realize I'm again I'm in the red dot Jeremy Smith (01:30:09.537) Yeah. Jeremy Smith (01:30:15.475) Yeah, right? Yeah, yeah. John (01:30:20.737) down at the bottom of low out of like however many dots, but still, I like not good nighttime reading. But then, um, then like the next day this other guy wrote this like, I personally don't even know if it was a guy, but there's other person wrote this, uh, you know, alternate version that was basically like, look, look, here's the last five tech technological improvements, you know, going back, I don't know, like 200 years. Jeremy Smith (01:30:25.718) Yeah. John (01:30:46.014) And every time everyone was like, no one will have jobs. And every time, if you look at it as a whole, not individually, but like as a whole aggregate, everyone's life got, you know, the aggregate life got better. and I was like, okay, that was a nice little palate cleanser to like wash the doom and gloom out. And that's what I've been everyone who asks who I talked to is not technical or everyone who asked about it. I send both of those articles because I feel like it's important to read both and to read one and read the one first, which is Jeremy Smith (01:30:52.247) Mm-hmm. John (01:31:15.917) like everything is. you're you're everyone's gonna be out of a job in six months, you know, like you read that one, you're like, geez, and then you read the other one and you're like, okay, like, maybe it's just some new normal, and I'm either okay with that or not. The thing I worry about the most are the people who just don't change well, there's always that, though that fraction of people, and you know that some of those people are going to be in your circle. And that's just really sad. So that's that's definitely the thing that I work worry about the most. Jeremy Smith (01:31:34.53) Yeah. Jeremy Smith (01:31:41.111) Right. John (01:31:47.465) But for myself, I'm like, you know, just again, like we've said before, like figure out how I can use it and what I can do and then and try and keep everybody around me aware as well and stuff like that. And then go from there. So. Jeremy Smith (01:32:05.536) Yeah, it seems good. I'm really glad that you've been, sharing so much about what you're doing, publicly. And, I think it's, it's been helpful to follow along and have, I dunno, like at this, these kinds of times, you like end up looking more for guides, cause we can't all be spending all of our time just experimenting or not that you're just experimenting all the time, but, the sharing the things that you're doing, like helps people level up faster. John (01:32:25.899) Yeah. Jeremy Smith (01:32:35.594) and figure out what's possible. that's a nice service. I've appreciated that. Yeah. John (01:32:40.587) That's good to hear. It's entirely selfish. Cause you know what else matters in the next? Well, it matters in the next world is distribution because building is no

longer the moat. So that's why I'm sharing because I'm like people buy from people. So the more I talk about this, the more likely someone's going to buy it from something from me. So again, it's purely selfish. mean, I like the stuff and I talk about it to everyone physically near me. I just was never good at putting it on, was not as good at putting it online. but yeah, it's, Jess (01:32:42.255) Yeah. Jeremy Smith (01:32:47.499) Yeah. Mm-hmm. Yeah. Right. Right. John (01:33:09.704) So that's why I was like, well, you know what? If I'm going to talk about this with everybody locally, I might as well just broadcast it. So that's that's literally all I did. But it is good to hear that it's helpful because sometimes it does feel like just flapping into the wind, you know, like so. Jeremy Smith (01:33:13.89) Right. Jeremy Smith (01:33:20.802) No, I hear you. I hear you. It's also there's so much noise right now around the subjects that it's hard to either feel like you can stand out from other things or just to take it in like to process it all. finding people that you trust and you know like want to hear their opinions that's really nice. So yeah. John (01:33:30.612) Yeah. John (01:33:48.064) Yeah, well, right back at you. I want to hear more from you guys. So it was nice. I mean, again, in slack and stuff, seeing some stuff from you, Jeremy and all that. So, I mean, Jess is just so heads down. I'm like, is he still alive? Just got to do like a health check every now and then. I assume you're working all the time. Jeremy Smith (01:33:53.984) Yeah. Let's go. Jeremy Smith (01:34:01.486) I barely get to see Jess. We know you're working. Yeah, we know you're working. Jess (01:34:01.86) Does he even work anymore? Jess (01:34:08.588) Yep, I gotta get back involved. It's definitely helpful. It's like every time I do though, it's like, you know, what they say, putting your mouth up to a water faucet. I just feel like I just, I just consume so much so quickly. I'm just like, whoa, I'm have to take a step back and like digest it all. Jeremy Smith (01:34:18.625) Mm. Jeremy Smith (01:34:22.793) Yeah, yeah, right. Yeah, easy to get waterlogged. John (01:34:29.12) I think the benefit for me is I've never been able to chug anything. So it just goes in and goes straight back out. and I think it just same. I'm very present. I'm just like in when I'm reading, you know, whatever Twitter, any of other things like it goes in and then I get excited and then it goes out. And I don't remember why I was excited, but I'm happy now, you know. But I think some of that's curation to like I like some, some people talked about how bad their algorithm is. So I don't know what makes mine good, but like mine is Jeremy Smith (01:34:33.033) Yeah. Jeremy Smith (01:34:47.349) Yeah, yeah. John (01:34:57.311) has been pretty interesting. And so yeah, and I think I'm just I'm into it and I feel like it's my job. Like there's you know, 10, 15, 20, you know, people at Boxout and across all the companies and stuff that rely in some fashion on me to not let us down. You know, it's not only me but Jeremy Smith (01:34:58.688) Yeah. Jeremy Smith (01:35:11.393) Mm-hmm. Yeah. John (01:35:14.673) I'm going to take it very seriously. And so I'm going to figure it out and I'm going to give the best stuff to them. You know, so like everybody has Claude accounts. Everybody has what and I tell them if you max it out, go to the next one. Like like that's use it on your free time on the weekends. Like not because I want you to work all the time. But like if you want if you build a totally separate thing and you spin it up and it does great. That's awesome. Jeremy Smith (01:35:20.853) That's cool. Yeah. Jeremy Smith (01:35:25.632) Mm-hmm. John (01:35:38.299) I'm not going to be mad that you did it with my cloud account or like a cloud account from a company that I'm a part owner in or whatever. Like my dad always had the farm guys take the trucks home, you know, cause he was like, I'm not going to make you drive a half an hour to the shop and then a half an hour back to a field. That's five minutes from your house. You know, like he had to do that. And he was like, that's stupid. So it's the same thing. Like I'm like build stuff, you know, if it works out great for you, that's fine. We'll figure out how to, you know, patch the hole. But, yeah, I, yeah. So Jeremy Smith (01:35:39.637) Yeah. Right. Through the company. Yeah. Jeremy Smith (01:35:48.127) Hmm. Let's go. Jeremy Smith (01:35:53.641) Yeah, yeah. Yeah. John (01:36:07.987) That's why I'm drinking from the hose. And because again, it just goes in and out. I don't actually get waterlogged ever. yeah. So. Jeremy Smith (01:36:11.381) Yeah. Jess (01:36:17.38) Well, that's great. Should we wrap it up there? Jeremy Smith (01:36:19.681) Yeah, I think it's good. Yeah, I can do a quick wrap up. We can say something. Jess (01:36:26.434) Yeah, I was going to suggest that. Jeremy Smith (01:36:27.936) Yeah, okay. Well, John, thanks so much. This has been a really good discussion. I knew we'd get into a lot of good stuff with you, and it's been fun. It's felt really natural. I hope it's encouraging to other people, giving them a vision for what's possible and how we can feel about it. That would be positive and good. So thanks for the time today. John (01:36:51.122) Yeah, thanks for letting me see your faces. That was great. I miss you guys. Jeremy Smith (01:36:54.209) Yeah, my guides. Jess (01:36:56.014) Been a good roadside chat. Jeremy Smith (01:36:58.601) Yep. Yeah, we appreciate it. Jess (01:37:01.656) John, anything you want to promote? John (01:37:05.158) I mean, you can feel free to check out any of the stuff that I'm involved in that I mentioned on this. But other than that, I just enjoyed talking and just want to promote get a Claude Max and go to town. So Jess (01:37:17.578) one quick thing. Are you gonna be in person anywhere this year? John (01:37:22.182) I have zero plans of being in person anywhere. We have a lot of camping trips, so it's going to be tough. Jeremy Smith (01:37:27.328) Nice. Jess (01:37:28.068) We're just gonna have to find you in a National Forest somewhere, Jeremy Smith (01:37:31.627) Yeah. John (01:37:31.815) You will. Yeah, that's actually happened. I ran into a Ruby guy in a net in. Yeah, no kidding. We were in Teresa Lake. It's in Nevada. There's a national park. can't remember the name of it now. I'm blanking. So I'll have to look that up later. But I was just walking and I was like, that guy looks super familiar. And I was like, do you know Ruby? He was like. Jeremy Smith (01:37:36.915) I'm not kidding. Jeremy Smith (01:37:47.627) Yeah. Jeremy Smith (01:37:54.209) That's awesome, yeah. John (01:37:55.96) Cause that's how I am. Yeah. And he was like, yes. And so then we chatted for quite a while and caught up and he was in like caboose back in the day, the IRC channel and stuff. So that's where I knew him from. So yeah, buffington, Michael, Michael buffington. Jeremy Smith (01:38:03.893) Huh, yeah. Wow, that's very cool. Okay, that's awesome. Jess (01:38:11.044) Awesome. All right, well hopefully maybe you'll get a wild hair and like come out to Asheville or maybe we'll see you in Austin or something. Jeremy Smith (01:38:22.507) something. Yeah. Jess (01:38:24.324) You never know. John (01:38:24.816) I some FaceTime. Jess (01:38:27.118) Thanks again, John. Best of luck to you and your businesses. Love hearing about them. Jeremy Smith (01:38:27.221) Thanks, John. John (01:38:28.68) Thanks.