Isaac: Welcome to Never Rewrite. I'm Isaac Askew.

Jeffrey Sherman: And I'm Jeffrey Sherman, and today we're gonna talk about the future of transient software. ⁓ so before we kicked off the show, Isaac was telling me how he's in a band and he needed software to track their performances and how much they suck. And ⁓ you did not, but I needed to liven it up. ⁓ no, they they practice, as you all should if you want to get good at something. And

Isaac: Hey, I didn't word it that way. Mm-hmm.

Jeffrey Sherman: They need to keep track of how they're doing with their songs. And so he vibe coded a piece of software with a single prompt and it solves this single thing. And now he's going to use it with his pract banned practice for their practices, but it's not a business, right, Isaac?

Isaac: Yeah, there's not much of a reason to make it a business because it's like a single prompt created it. So anybody who wanted to be at my competitor, we just prompted themselves, you know.

Jeffrey Sherman: Mm-hmm. when you described it, here is how I in my mind I would build such a thing if I was going to build it manually in the old school way, right? I'd probably use Google Forms to input my data and then I would have see then you'd have to have some kind of visualization. ⁓ So okay, well I'm back our I'm immediately back to ⁓ I need a you know a three-tier crud app where there's a database of songs and how many times we performed them and how we feel about them, and a a dashboard that says these are the songs, this is how we feel about them, and then some kind of input the page that says, you know, this is the practice date and this is the times we recorded them and this is how we feel about them.

Isaac: Yeah.

Jeffrey Sherman: would still take you ⁓ a day to build all those things, you know, over time, and you'd probably have to spend ten a month to keep all those services running on the cloud. So how what it what under the covers, right, so y you prompted it and it wrote all the code. How does it work? Like where does the data live and yeah.

Isaac: Well, let me first start with the problem, I guess. What we had done is we had like a Word doc, and I just shared the Word doc with everybody, and we would all go on the Word doc and like add song ideas that we wanted to cover. I like, kind of, I found out like there is a voting feature for people, like you can suggest something and people can vote on that, but it's like sold on a particular type of Google doc. account, like the education account versus the business account. And I had access to the business account, but not access to voting through that. And I realized they were like, Google has done some kind of thing where they've hidden some really useful thing that would be nice to vote on an idea or like emoji thumb respond suggestion or something like that. And so I got kind of annoyed with them. Like, man, like I'm trying to nickel and dime for all these little things. ⁓ So we would just go in there and update it with like a number next to it or just put our initials next to it. Like this was very manual.

Jeffrey Sherman: Hmm.

Isaac: Like we like this song. And then we'd have another tab and the tab would be here are the songs that we're ready that we're gonna have in our full set list and some other songs that we just added and songs that we, you know, we might have like a little note here and there or comment that's like, not, know, the drummer doesn't have this one down yet or one of the guitarist hasn't learned the solo for this one yet, something like that. ⁓ And so not ready to perform. And I would just, I would send everybody a group text. Hey, can you go in there and add some songs or

Jeffrey Sherman: Mm. Not ready to perform.

Isaac: Is everyone good with this? And sometimes people would do it and sometimes people wouldn't, I couldn't really tell. And then, you know, after a while, I was like, you know, I think I can just vibe COVID out for this. you know, it didn't, it wasn't until it became, until like this friction point of like, okay, I'm not getting feedback from some people. I don't know where everyone's at on the new ones. It wasn't until we really added a whole new series of songs, we actually got invited to play a full set. So we don't suck, Jeffrey. ⁓ And so we didn't have enough songs to fill the set. And so we were like, let's add a bunch of songs. And then that's when we really needed to start tracking the progress because we're like, we want to fill up an hour as soon as we can. Where are we at? ⁓ So I took a very small prompt. actually had took the word doc and I like command a, you know, selected everything, copied and pasted it into the prompt. It was like, this is how I'm managing.

Jeffrey Sherman: Yeah.

Isaac: my band currently, these are the songs, can you make an app that shows our set list, also the set list time, like how many minutes have we covered with the songs that we've got so far, which would be useful to track, and a way people can check in and give status updates. That's it. I didn't give it any details, I just thought, let me just let it run with it and figure out what it thinks that I might like from this. And it like one shot it. It was...

Jeffrey Sherman: Hm.

Isaac: Excellent. They did everything we needed. There was only one bit of feedback that the bands had, which was like the rating was one through five to begin with, and they wanted a zero, which is I haven't started to learn this one yet. Whereas one was like, it's coming along. I'm like, OK, whatever. Easy enough. I can add a zero, you know. And yes, I can see when everyone's checked in which songs they've said are coming along, which aren't. And then it ranks them. It has like a, you know, either green to red and

Jeffrey Sherman: Okay.

Isaac: the red ones are the ones that are, you know, some people are struggling on the number of times we practice them. And it floats the ones up to the top that people have the most disagreement on essentially. And then we used it on Monday for our first practice on which ones to prioritize ⁓ very successfully. And then later in and added some bells and whistles. I was able to add like for the songs that we're learning, you can click and it'll take you to ultimate guitar for the tabs or clicking. It'll take you to YouTube, but the actual song, you know, if you wanted to listen to the live song again,

Jeffrey Sherman: Hmm.

Isaac: play around with it. So little things like that. But yeah, there's not a reason to sell it because again, like a single prompt made it. I don't really have a moat to defend. And it's very personalized software for our band. Like I don't, you know, it solves our problem and there might not be a problem other bands have or they could be other software people use. But I certainly don't need to pay for the Google education version to get voting or pay for

Jeffrey Sherman: Mm-hmm. Right.

Isaac: any of these other things to text people to let them, hey, check in here. I don't want to pay for these little things. I'm already annoyed by all the subscriptions I've gotten, all the Netflix and HBO. I don't want to pay for another thing.

Jeffrey Sherman: Yeah. So I assume right you've got your own consultancy, so you kind of stole your own office supplies and put the tokens ⁓ you use the tokens on your unlimited plants. No.

Isaac: No. No, I'm not stealing my own. But I do have two different plans. I've got for YGG, my consultancy, I've got Codex, which is like the 20x plan. It's $200 a month. I have the exact same or the analogous one for, oh sorry, that's Claude. And for Codex, I have the same one, which is like the $200 a month one that's like essentially the same thing.

Jeffrey Sherman: Okay.

Isaac: And so just use one for my personal stuff and then one for the my work stuff. And it lets me also compare the models too, so I can see which one, because each one releases new models and they're all fighting each other. And everyone's like, ⁓ Opus is great and Fable's great. And some people are like, ⁓ you know, the new Sol and Luna are great. And then it lets me test out how much Fable burns token wise, because I can feel it rather than just hear people complain about it. So I use both. And in this case, this was a personal project that I'm not using. So I used to use my personal account for that

Jeffrey Sherman: That's good 'cause I as as soon as I said it I'm like, ⁓ but then I bet your so I'm that it creates a gap so that your personal stuff doesn't bleed into your work stuff and vice versa. It would be much worse if your customer stuff bleeds into your personal, yeah.

Isaac: Yeah, because also it'll- It'll learn from what you're doing. Like from a personal one, like I mentioned, what did I mention the other day? Just something random. And then it pulled for, you know, and this will also pair well with the basement studio you have set up. And I'm like, oh yeah, I talked to you about my basement studio in a different thread. You're right. And so it'll, it'll still, even if it's not related and that's like an archive thread, it will still remember some stuff that I've told it. So I try to keep everything separate so that it doesn't bleed in and start mixing suggestions in. Cause

Jeffrey Sherman: Mm-hmm.

Isaac: It does some weird magic behind the scenes.

Jeffrey Sherman: Okay. So you've got your own ⁓ personal ⁓ account for the tokens and then how infrastructure wise, how does this app work? Is it actually like an iOS app or is it a web page? How's this thing like running?

Isaac: It's a web page. it built it. Honestly, I don't even know if it was React. I didn't really care, but I gave it the ability to deploy via Cloudflare pages. I already had it connected for, I ported our old site separately on the work account for Never Rewrite and for YGG, which was using Squarespace at the time. And I had it port that to Cloudflare. And so I had the new one do a similar setup where it supports the Cloudflare or connects to GitHub. And so anytime you push to GitHub, it'll automatically deploy the site. So the other ones, I think they're built, some flavor of React, I forget. But it does all the config for you if you have connectors set up or clients set up. So once you actually establish that, you can just be hands off, it'll push domain and then it gets deployed automatically. And then the key value store, at first it was trying to use SuperBase, but I'm like, I don't wanna use SuperBase for this. So I ended up using some other kind of key value store that was a lot cheaper to manage, but I even, forgot about that, which kind of leads us back into this other concept of it, iterating and building stuff that we don't even know how it works, which is fine for me, cause I don't really care how it works as long as it solves my problem.

Jeffrey Sherman: Yeah. Right, if you get hacked the the I guess the worst thing that could possibly happen if you get hacked is that somebody could mess with your proposed set list for your next gig and

Isaac: Yeah, it is password protected. I just gave the band like the actual password to get to the web page. ⁓ But yeah, if they hacked it, they would hack with the number of times we ran through a setlist. It doesn't need to be, you know, anything special anyway. So again, it's transient software that I'm using just to solve the problem. I came up with that. I'm sure that maybe it already exists as a thing, but I'm just calling it transient just because it's like this temporary floaty kind of software. And if the band dissolves, No one uses the app. I'll just spin the app down and you know, it doesn't matter. Like so software, things that might have taken a long time to build and you cared more about and care about the deploy and the costs. And this Cloudflare is the pages thing is it's free for the amount of traffic we've got. And so it was like every part of this thing and I'm ⁓ under the $200 a month plan for that for Codex and I haven't used my tokens for this month. So I'm like

Jeffrey Sherman: Right. Hmm.

Isaac: From my perspective, since I've already used it for other things, then it's somewhat free for me to build this on the side, because I've got this thing I've already paid for, monthly fee, generating some code, hosting in a thing that doesn't cost me any money. No harm, no foul. I can see this being, especially because I, technically, I didn't, I mean, other than like some knowledge of what Git is and what Cloudflare pages is, which I guess is a little bit of a learning curve for somebody who's not.

Jeffrey Sherman: Right.

Isaac: into tech. Other than that, like this required nothing for me. Like I said, I don't even know really how it's built. I just know that it built it and ran it locally. It looked good to me. I said, yeah, go ahead and launch it. I don't care how it's built. I don't feel bad that I don't know how it's built. I just know that it works. And I feel like anybody else who gets to build software like this, they can actually, like I could give this to any of my friends ⁓ and say, hey, this is what I did and say, here's this tool.

Jeffrey Sherman: Right. Well

Isaac: And I'm sure they could walk through and discuss and it could tell them how to set up Git, how to set up all this stuff and they could do the same thing.

Jeffrey Sherman: Right, I was just thinking like as soon when you said, ⁓ I could t give it to my friends, I'm like, I bet you couldn't. Like, I bet if you tried to say, ⁓ I've got a friend who's got a band and I explained it to him and he thought it sounded cool, I bet setting up to have it have two instances would be a disaster. ⁓ where it it would require some rework and I bet you would spend up more time than you expect so that you could have multiple instances deployed off of the single Git repo.

Isaac: Hehehe. Oh, I wouldn't try to do multiple instances. There's no point. I don't need to copy my software. All I to do is let him know, hey, talk to chat dbt to understand how to set up cloudflare and how to set up git. And then just ask it to make an app for you. And maybe it builds it into a completely different technology. Who cares? You know, like I don't need to duplicate my code anymore. then, or it, cause it also duplicates my solutions to problems that I have. He could have a whole separate set.

Jeffrey Sherman: Hmm. Mm. Right.

Isaac: You know, so it's also like as part of the whole AI slop thing, right? Where you have a ton of apps on the app store. They're very hyper personalized. I'm sure like the number of workout apps has quadrupled or more because everyone's like, I don't want to pay for this dumb thing. Cause I think we talked about this in one of the prior episodes where you can.

Jeffrey Sherman: Right.

Isaac: make like I had this one called strong and I could log my workouts, but I couldn't get like a history, like a report of how it worked over time unless I paid. That's how they got me. I'm like, I'm so made my own app. All it is is logging the reps and then let me track over time. I could make that, you know, and you can see also like if everyone's publishing their own thing and they go, Ooh, I wonder how I can make money off this. Then you're back in the same loop.

Jeffrey Sherman: Mm. Right, it it's an interesting thing because because the software is so cheap and it now that it's transient, right, it won the stakes are much so much lower that you don't have it's not that security becomes unimportant. It's just the if somebody hacks you, it's not the end of the world. It's like ⁓ in the old days ⁓ and I'm talking nineties, where eventually your hard drive would die and you would lose everything.

Isaac: Yeah. Mm.

Jeffrey Sherman: And so you would make sure that you had some kind of backup and and you you'd go over your friend's house and say, look, my hard drive's making a different noise, it's gonna die soon. Let's go back up all the things. And you get out the floppies. And that's fine. But and eventually it would die. Just be like, all right, you be like inbox zero, I've just deleted everything. And it's like it would take you a day, you'd set up a the computer, a new computer, and you'd be back in business. And it's just kind of like that, where it's like, well. I don't have to pay anybody, it's what I got and move along.

Isaac: Yep. Yeah, just the fact that it's so easy to spin it up and generate this code. I almost think you don't want to commit it to get like I did. You're just adding more crap. This one actually can live on my hard drive. And if I lost it, I'm like, ⁓ Hey, can you make that again? Find the archived process and say, hey, can you make that again? I lost it. Because it's that easy. Now, it's expensive now, and I'm sure that's not, you know.

Jeffrey Sherman: Mm-hmm.

Isaac: Well, I say expensive again, I'm underneath my monthly plan. So don't really know how much I've spent technically, because I don't pay by token in that sense. pay by the weekly usage or whatever in the model that I've chosen. But I could even see this. mean, I have played around with like a locally hosted LLM too. And I imagine at some point, those will be good enough too, to where I could just host my own locally, ask it to build it. And that really is just whatever the cost of my electricity is to produce the thing.

Jeffrey Sherman: Right, and in this it kind of goes to a problem that's ⁓ the problem of bloatware, where if you've got a SAS and you've got a subscription and they feel like they need to put ever n they need to continually give you new stuff to justify the cost. Also s you know, so becomes this ⁓ a lot of software trips over into bloatware where the cost is rising because they keep hiring developers to add features that you don't care about so that they can justify the rising costs.

Isaac: Yeah. Mm-hmm.

Jeffrey Sherman: And in this case it's like look, I I want a specific piece of software and you'd pay twenty bucks for it, but any more than that you'll just vibe code it because you you wanna pay once and then you're done. And it doesn't need to be perfect. It doesn't right. That's right. That that

Isaac: You're right. 20 bucks one time. Not paying $20 a month for anything anymore. As an engineer, will find a way to, like I have so many crons running locally now doing things, because I'm trying to avoid paying for these $5 a month here, $10 a month here services. I'm like, if I pay $200 a month for this maxed out plan, I am creating all kinds of software that's going to solve my problems here. Then I'll just cancel that for one month and go to, $10 a month plan or whatever, $20 a month, something cheap once I've got most of my stuff Vibe Coded.

Jeffrey Sherman: Yes. It it makes sense. ⁓ so yeah, that's the future. Get out of the SaaS business, I get. Well, unless 'cause we did an episode of this as well of it software can become transient but reputation cannot ⁓ and this goes yeah, we we talk about things like mail servers where it's not about the software, it's about the reputation.

Isaac: future transient software. Right. Yeah, I can't vibe code that. can only vibe. I'm vibe coding simple things. In fact, anytime I come across dumb stuff like, ⁓ Google built a feature where they can thumbs up a comment in Google Docs, but they removed it for everybody who doesn't pay for the education plan. That's when I go, hmm. Screw you. Out of spite, I'm going to use my tokens to make my own version of this. I hate this. So I'm hoping that makes Google change their mind about what they're stingy about. I think it'll lose them money over time. We'll see.

Jeffrey Sherman: Yeah. Probably not. Well thank you all for listening. I'm Jeffrey Sherman.

Isaac: I'm Isaac Askew, and this is Never Rewrite.