Vibe coding

Rick (00:00.462)
What's up Tyler? Oh, you're a new dad.

Tyler King (00:02.48)
Hi Rick. We were, right before we hit record, we were both like, we're super tired right now.

Rick (00:10.862)
Well, how are you doing? You're a new father.

Tyler King (00:13.808)
Yeah, I've procreated, I've achieved immortality, so yeah, doing good. I'm happy to dive into more detail, but overall I think it's been about as smooth as it could go.

Rick (00:28.27)
Was it everything you, were you prepared, I guess, mentally? Or has it been better or better than you thought, easier than you thought?

Tyler King (00:37.778)
Probably about the same. Yeah, it's a lot, right? There's just a lot of stuff, but it's funny, all this stuff, you I tried to prepare as much as I could. You try to read books and have all this stuff and all that, you know, you can't really prepare, as you know. But I have been pleasantly surprised by how quickly things...

become normal. Like even just holding a baby. I've never really felt comfortable holding a baby and I was just like, no problem. And you know, just, changing a diaper and this and that. So I would say I wasn't exactly prepared but I've adapted well enough and I know there's just, we're gonna keep getting punched in the mouth over and over and over again but I think we're okay where we are right now.

Rick (01:23.95)
That's awesome. Congratulations.

Tyler King (01:27.782)
Thank you, thank you. The reason I'm tired is our routine is Shelly's mom is staying with us. so her mom, yeah, Shelly's just like doing the things that only a mom can do. And then her mom is like the support role during the day and I'm the support role at night. So I'm doing the night shift. So it's 9 a.m. right now. I'm kind of coming off of my dad duties and I'm gonna go to sleep right after we record here. So that's why I'm tired.

Rick (01:36.595)
that's nice.

Rick (01:51.534)
yeah, Yep, yep. you, well, is it too early to say it's time for number two or are we done? Okay, that's awesome.

Tyler King (02:03.874)
I think if you said that to Shelly right now, she would break down. Yeah, you know, there's a physical toll it takes on a mom that it does not take on me.

Rick (02:12.962)
Yep, yep, it's crazy. But hey, we're right there with them. We're there.

Tyler King (02:18.29)
yeah, totally equal, totally. Getting to take parental leave is so nice though. So, Shelly's mom and stepdad are here and they're just from a different generation and they're like, the dad didn't get any parental leave and the mom was like, I think like two weeks or something. I don't know how anyone did it like that.

Rick (02:34.04)
Hmm.

Rick (02:43.116)
Community, it wasn't, I think it was a lot less family sort of handling it themselves, but like neighbors and communities helping out. I think that's what I've heard.

Tyler King (02:44.517)
Yeah, that's fair.

Tyler King (02:55.216)
Yeah, that's probably true. Can I just say one of the things that I find so funny about like parenting in the modern generation, like one of the norms is people offering to like bring you food or whatever. And some of those people even listen to this podcast and I'm very appreciative of everyone that said that, but it's just like guys, DoorDash exists. Like I'm totally fine on food.

Rick (03:15.374)
Yeah, it's like that being, you know, like a huge burden lift is like not existent anymore.

Tyler King (03:25.596)
Yeah.

Yeah, if we had to go to the grocery store and cook, that would be miserable, but it's like, I'm I'm eating well. I'm just getting Taco Bell, Wendy's, Chick-fil-A. I'm eating everything I want right now.

Rick (03:38.83)
Well, how are you spending your, let's see, how long has it been since you came home?

Tyler King (03:45.458)
This is day nine. so came home, yeah, seven, six days ago.

Rick (03:49.182)
gosh, yeah, it's... So you're kind of getting into a routine now, which is good.

Tyler King (03:55.92)
Yeah, yeah, she's backsliding on the feeding part, but yes, aside from that.

Rick (03:57.902)
No, that's scary. The scariest part is when they put the baby in the car with you and you're like, I'm responsible now.

Tyler King (04:06.672)
Uh-huh. I've never driven so carefully in my life. Also, normally, like Shelly's normally the driver. Like I didn't even own a car until a couple months ago. and, but she can't drive right now or she's not supposed to. So I'm like doing a thing I'm very unfamiliar with, with like this fragile little child in the back. Yeah.

Rick (04:11.456)
I'm sure.

Rick (04:24.878)
Precious, precious, you're two most precious people. So, well, I was just gonna ask, how are you thinking about work? Do you have more time than you thought, less time? How are you gonna spend this time?

Tyler King (04:38.802)
Uh, yeah, so I mean, I've, do I have more or less time? Probably, I have plenty of time, but I'm tired enough that I'm not getting, making good use of it yet. But you know, we're nine days into a 28 day parental leave. So I think it will just get better and better, hopefully. Better as in like, she sleeps for longer chunks of time and that's, and I can sleep. So, uh, cause I have plenty of time, but I'm spending that time just like zonked out.

Rick (04:49.212)
Mm.

Tyler King (05:08.498)
I have gotten some work done, you've seen. I've been doing like up health stuff primarily. So I've got, I mean, I want to talk about that a little bit as a topic later, but yeah, I've gotten half of what you want me to get done, done. So I think I'm pretty definitely going to get kind of phase one of what I'm working on done during this leave.

Rick (05:08.919)
Mm-hmm.

Rick (05:12.344)
That's awesome. Yeah.

Rick (05:29.39)
That's awesome. That's awesome. Yeah, for me it was like after about I'm interested, I'll be interested at the next podcast. You're like, I'm ready to go back to work. Or if you're like, this is great. I want more time off. That's going to be like for me it was like two, three weeks. I was like, okay, I think I'm better served going to work than sitting around here.

Tyler King (05:39.579)
You

Yeah.

Tyler King (05:47.694)
It's time.

Yeah, there's only so much you can do. Yeah. Cool, but yeah, that's me. What's going on in your world?

Rick (06:00.032)
Kind of the big thing that happened this week is Windfall, my day job, announced a pretty significant fundraising milestone where we raised $65 million from Morgan Stanley Expansion Capital. Kind of a big deal for me because I joined right after the Series A about three and a half years ago. So to kind of help the business get to the stage and be like somewhat, not like a, mean, the founders drive a lot of the outcomes.

But to be a part of that executive team to get to that milestone was really like kind of rewarding, not in terms of necessarily like just the financial piece, like the just like, oh yeah, like a third party validated the value created at the business. And it's kind of like time of reflection for me. I've just gone, oh wow, like this was a meaningful journey. And then also like I've learned so much.

I can't, my skill development, I didn't, you kind of forget, I don't know, forget what the, I think it's called the planning fallacy or something similar to that, where it's like, you think like you've changed all you're going to change. But if you like, if you look forward, you don't think you're going to change. But if look backward, you're like, oh, I've changed so much. I've changed so much over the last three and a half years in terms of my, especially skill development, technical skill development, but also like management and leadership skill development.

Tyler King (07:16.784)
Yeah.

Rick (07:25.646)
And I just, I've taken the time right now to reflect on that and appreciate it. And it's great. But it's kind of back to work today.

Tyler King (07:31.526)
Yeah, congrats. Yeah. Yeah, that's a funny thing about like fundraising milestones is it's, you know, in the news, it's as if that's the end and it's, you know, that's the beginning. Yeah, yeah. I find it so funny. sorry, good. I find it so funny how like how much of a juxtaposition you've got of like the startup to last.

Rick (07:46.324)
Yep, yep, beginning of a new journey. So, super, super excited. Go ahead.

Tyler King (07:58.566)
side of things and then like, we just raised $65 million.

Rick (08:02.033)
Hey start the last companies can raise $65 million. Okay.

Tyler King (08:05.234)
No, I disagree. There's absolutely no path that leads to sustainable, just chilling out, calm company if you've raised 65 million from VCs.

Rick (08:18.786)
Yeah, if you've raised any money from VCs, it's very difficult.

Tyler King (08:21.554)
Yeah, there's the occasional story of like a Gumroad type thing where you raise money from VCs and you basically fail. The VCs give up on you and then you buy the company back or something like that. I think at 65 million raise, and I assume so your valuation must be much higher than that, it's kind of boomer bust at that point, I feel like.

Rick (08:41.098)
Yeah.

Tyler King (08:43.874)
I'm not saying that as a criticism either, like if you want to, you know, become a billionaire, that's how you do it.

Rick (08:48.312)
Yeah, no, and it's like exactly right. So like up health is your point was they're very different, mindsets. and I don't think you could, like, think part of that's why I can progress both both at the same time is if they were two of the same, I think they'd be very competitive with each other. in terms of like the type, like the brain power, but they're very, it's actually like leg up health is almost a relief of, you know, windfall, and then windfalls, a relief of leg up health in terms of me not.

being a crazy person. I could be crazy at windfall.

Tyler King (09:21.712)
Yeah, well, and yeah, I'm definitely curious if, you know, this would be far in the future, but if one day you're not just full-time on LikeUp Health, but if you were full-time on LikeUp Health, you'd get bored with the calmness right now. Like, the calmness is good for you as a very fractional CEO and stuff, but it'd be interesting to see you apply all these hyper-scale skills you're building to LikeUp Health and see what, like,

That's where that could take us.

Rick (09:52.054)
It would increase the change, like pace of change, which could be good or bad, I don't know. It would definitely.

Tyler King (09:59.92)
I mean, I think the main thing it would mean probably is, certainly me, but probably JD also get, like when you hire your first employee, it's tempting to be like, this person's gonna be, you know, I'm the head of the company and this person's the second lead of the company forever and ever and ever. And that's not normally how it works. Normally like the first hire is more of a startup entrepreneurial type person. And then as you professionalize, that person either gets.

Rick (10:13.326)
Yeah.

Tyler King (10:24.796)
they leave the company or they find some little niche role or whatever, it probably affects how JD and I interact with the company quite a bit.

Rick (10:31.21)
Interesting yeah, I haven't thought through that yet. That's scary Yeah, that's scary. I don't want to talk about that

Tyler King (10:33.722)
Yeah, this is a pointless thing to discuss. It's just been... I actually have sort of that, not with JD, but with me, sort of that as a topic for today. This wasn't maybe my most important thing, but since it segues nicely, you want me to dive into that? So, well actually, let me do both my topics, because one leads to the other. So I've been working on this LegUp Health stuff, which, just to summarize for the listener, so LegUp Health...

Rick (10:49.624)
Yeah, go for it.

Tyler King (11:03.708)
primarily a health insurance agency where JD is providing the service and that's the main value we're providing. I built some software so that employers can add their employees, but the software provides almost no value right now. And so my project right now is to actually give people a little more value from the software itself so we're more of a tech enabled service instead of just a service. Any modifications you'd make to that description? What are you laughing at? Just the...

Rick (11:28.238)
No, that's fair. Just the blunt honesty. I don't want to admit it, but it's true.

Tyler King (11:34.962)
Yeah, every time I log into it, I'm like, what is this for? But it could provide value in the future. So what we're adding is...

Rick (11:41.358)
Yes, yeah, yeah. Here's what I'll say just to make sure it's not like snake oil is is is the point here is it is a interface right now it is an interface for employers to tell like up health about people they've hired people that they've given money to and people that they've let go so that we can serve them with their consumer health insurance like that's the point.

Tyler King (11:49.729)
yeah.

Rick (12:10.452)
And it has the, based on our initial adoption and customer feedback and how people are using it and also JD's pipeline sort of like sales learnings, as well as our historical experience at other companies that are relevant. We, we, know that like, there are like a lot, there's a roadmap here of, of development features that turn this into much more functionality and value for that. And then just like telling like upheld about employee statuses.

Tyler King (12:40.134)
Yeah, so this, like what I'm working on right now is kind of the first step of this, is prior to this recent project, when the employer adds an employee, they can like add an allowance, which is basically saying, I'm gonna give this employee this amount of money each month to help pay for their health insurance or whatever. But we don't do anything with that allowance, even still in this new thing we won't, like we're not moving money around or anything like that. It's just to tell the employee what their allowance is, right? What I'm adding is the ability to make multiple benefits programs to say,

here's your health insurance reimbursement thing, here's your cell phone reimbursement plan or whatever. I don't actually have good sense of what the different plans will end up being, so that the employer can kind of build a benefits program that's more than just, hey, give us your health insurance information and JD will be your agent, right? But it's still very much a basic CRUD SaaS app. So the one thing I wanted to mention, I've just been vibe coding this all the way.

Rick (13:38.296)
What's vibe coding?

Tyler King (13:39.954)
I love how offline you are, Rick. Vibe coding is, it's the new thing everyone's talking about related to coding with AI, and especially people who don't know how to code using AI to do it. Or actually, you'd be a perfect candidate for vibe coding, Rick, because you know the coding principles, but there's a lot. You're not practically a software engineer. Vibe coding is just like, ask chat GPT, or I'm using Claude, like, hey, what do I do?

Rick (13:41.838)
Okay.

Tyler King (14:09.97)
and it just tells you and then you do it and then you're like, no, that didn't work. And then it's like, oh, okay, here's why. And you just talk to it, but you never really understand what it's doing. You just copy and paste code in that it gives you. Now, obviously I have a stronger foundation in coding than that, but I'm trying to move really fast and get a lot done with a limited amount of time. I'm definitely, I'm like, here's the current file. And I just tell it like, here are the 10 things I want changed about it.

give it to me and it gives it back. copy and paste it in. I test it. I'm like, nope, here are three bugs I found. Give me update and it just gives me the updated code. I'm not like writing any of the code really. Yeah, it's weird. Your face looks like your mind is blown.

Rick (14:50.638)
Okay, that's awesome.

Rick (14:58.456)
Well, I'm just like, like, what does this make like the code like more like

like risky I guess like is the word or less scalable I guess maybe I'm turning I don't know what the right words are.

Tyler King (15:09.808)
Yes, I have things to talk about with just the vibe coding in general, because I used AI to build this originally, AI wasn't as good, but I think it's less that. It's that I didn't understand how to use I didn't trust it as much at the time. So this is a different experience for me. So I have some things to say about that, but to skip ahead, because this is what the segue, this is very long segue, in terms of if the company scales, what's my role and stuff.

This is an incredible way to get, I'd say like what in the back in the day would have been more like a full-time developer's worth of productivity out of me, much less than a full-time developer. The flip side of it though is it's gonna come with a lot of, it's gonna look like I'm just really productive. And that might be tempting. The reason I'm saying this is to set realistic expectations. It might be tempting to be like, with the help of AI, Tyler is as if we had a full-time developer and like let's have

ambitions as if we were more of a tech company and not just this, you know, a couple hours a week. And it will work with very simple cred stuff like this. But I'm also seeing how fragile the foundation is going to be in terms of adding like more complex stuff to this.

Rick (16:25.304)
Like what, what, what gets, what, what do you start thinking about when it comes to complex stuff? that like integrations?

Tyler King (16:32.1)
Yes. So definitely, whether I'm using AI or not, integrations, moving money around, anything that could involve money calculations, like we're collecting the employer's allowance right now, but we're not actually telling them, pay this person this amount. As soon as we're like taking, you know, claims from the employee and calculating a balance, at that point, you can't be quite as fast and loose as I'm being right now. Yes.

Rick (16:58.584)
Yeah, you gotta be precise.

Tyler King (17:01.076)
And integrations are just always hard.

Rick (17:03.07)
So what you're saying basically is it's less about necessarily the technical complexity. It's more about the precision, the room for error. Yeah, yeah. Yeah.

Tyler King (17:11.612)
I think so, I'm just guessing, because I haven't hit the actual failure mode yet. But like, yeah, I've got a couple files that are, I wrote the code before and it was, let's say 50 lines of code. And then I gave Claude that 50 lines and I said, here's what I wanna change. And it gave me back 600 lines. And I'm like, this does not need to be 600 lines of code. This should be 60. This should have added 10 lines of code to what I had originally. And I like glanced through it, I'm like,

Rick (17:29.526)
yeah.

Tyler King (17:39.962)
It's not that anything it's doing is bad, it all works, it's not wrong, it's just...

Rick (17:44.782)
Is it easy to like look at it and go, this is what's happening or is it harder to do that than what your code would have?

Tyler King (17:50.49)
It's definitely harder than my code would have been. But the thing is, here's the magic power. Claude can look at it and tell what's going on, right? So as long as I'm comfortable saying, hey, here's the code you wrote before, help me make it better. it does, I wouldn't wanna work on it as a person. And I've been giving it a lot of feedback. Like a lot of people who doing this are, know, kind of what you were when you built the no code version, but they would do it with vibe coding instead.

Rick (17:57.592)
Hmm.

Tyler King (18:19.088)
But I'm looking at it, I'm like, okay, you just wrote the exact same thing 10 times, turn that into a common function, like what are we doing here? But if you didn't know anything about code, you wouldn't know to tell it that, you know?

Rick (18:25.806)
Yeah, yeah, yeah, Soon, I anticipate AI will come up with that. Would you like us to optimize this code for you? Like, that's a natural evolution of AI, so I worry less about that.

Tyler King (18:40.048)
Yeah. Yeah. And it works fine. And I can test it. in many ways, it's actually like one thing I've never done. Do know what unit testing is? Okay. So I've never, I'm just not someone who has, I'm too lazy to do that. But with AI, I'm like, hey, right, you know all the code, write me some unit tests. So in some ways, it's actually gonna be better tested and more robust than if I were to do it myself. So anyway, on the one hand, I'm very excited that I'm like, I think I can build some good functionality.

Rick (18:49.203)
Yes.

Tyler King (19:09.68)
Rick (19:09.72)
You can build more stuff faster with less time. The cost is one, don't think this is scalable. And two, let's be realistic about the long-term scalability of this. And then three, be wary of complex things. Or it's not the complex, it's more like things that require accuracy reliably and can't fail. Those things are going to take longer than the other things.

Tyler King (19:37.734)
Yeah, because I would still need to write those by hand. And like the thing I theorize would happen over time, if we keep building, so why am I bringing this up? We're gonna need, I'm very confident I'm gonna get this done during my parental leave. I'm not gonna probably, I consider this probably the bulk of the work I'm gonna get done this year, but then at some point we're gonna be saying, what's next? It's gonna be tempting to be like, okay, well let's just do simple crud stuff over and over and over again.

Rick (19:39.95)
Yep. Yep.

Tyler King (20:05.734)
But even that, I worry that if the footprint of the software gets too big, like one thing that AI is not good at is understanding the full context of the code base. It's good at saying, here's this one file, can you do this thing to it? But it's bad at saying, let me ingest the entire code base and understand and add something that is gonna touch all these different files. So even if we stick to the really simple stuff, if the footprint of the code expands too much, I think that could be a problem too. So one of the things I'm...

Rick (20:13.23)
Yeah.

Tyler King (20:32.837)
I don't have my head fully wrapped around this, but I think I might advocate for like, thinking, is there stuff we can do that stays within our current footprint a little bit better instead of just piling on more and more and more outside of it?

Rick (20:46.252)
What's an example of in footprint versus out of footprint?

Tyler King (20:49.478)
Yeah, so like one thing I was thinking of that might make sense, anything that kind of cleans up the process for JD. So for example, building billing. Right now JD just like manually, think what he, I don't know, have no idea, but I think what he's doing is each month he's going in and seeing how many users each account has and updating that in Stripe. Like getting that pushed to Stripe, or just like building a billing system. In a sense, it's complicated and it is an integration and all that, but it's like the same.

core functionality we've got. It's the same database tables, it's, you know, versus being.

Rick (21:23.054)
It's making it more fat. It's making the existing service faster and more efficient versus adding new service.

Tyler King (21:29.552)
Yeah. Or like, let's make the email we send someone, like when you add an employee, we send them a very generic email that doesn't say anything about their benefits. Let's update that email to mention the specifics of their benefits. things that, yeah, enhance what we've currently got rather than being like, okay, now we've got an HRA and now we've got claims processing and now we've got this and this and this and this. We could do that. I just worry about the fragility of the software if we do that.

Rick (21:55.81)
Yeah, it makes sense.

Tyler King (21:58.546)
But yeah, it's cool. mean, you can just go in and, again, I am looking at the code a little bit as a hobbyist sort of, but I think you could do this without ever understanding any of the code.

Rick (22:12.782)
and get it functioning out.

Tyler King (22:15.184)
I think so, yeah, this is, let me be clear, very much for prototyping. don't, like you couldn't build less annoying CRM this way, absolutely not. But because like upheld's relatively simple, like yeah, I think you could.

Rick (22:22.936)
Yeah, yeah.

Rick (22:28.622)
That's amazing. Um, I have a chat, have a, um, chat, GBT sort of when to, share for leg up health. Um, I was on the plane and there was like, so we had, so I went to North Carolina, um, which I guess one thing I wanted to say is my family lives there. I haven't gotten to see them a lot since we started having kids. Um, and the Avery and Oliver just got to the age where it's like, oh yeah, we can go on a plane now and a big deal. Um, well, it wasn't a big deal. It was a hard flight for a half hour flight, but, um,

but, but I, always, every time I go home and we spend time quality time there, I come back so calm and re centered. I, just makes me wish my parents and siblings were out here, in Utah, but it's all, it's all good. but anyway, on the plane, I had like this hour and a half of like, okay, you know, our niece is sitting next to me and then I've got my, you know, two, two kids on the other side and they're just like watching movies. So I pulled up my laptop. I was like,

I remember JD asked for an HSA guide, you know, three months ago, I'm going to do that for right now. So I started writing that. I was like, let's see what chat GPT come up with. Dude. The first draft was like better than any draft, any content writer I've ever hired in the history of my life has ever come up with like in terms of the content. And then, so I pulled that into a Google doc and I started editing it and I'm like, okay. Um, wow. There are like 30 other blog posts in here that I could link to. Let's let's build drafts for those. So I basically took the.

Tyler King (23:41.862)
Yeah.

Rick (23:55.106)
I basically kept my same prompt and then said, can we write a sister article answering this question? Can we write a sister article on this topic? And I did like 15 of those. In an hour and a half, had first drafts and edited first drafts of probably 20,000 words to 30,000 words, a full eBooks worth of content, 15 articles. And I guarantee it's better than most content out there that you would find, because I'm editing it.

And organize it's like, like you can't ask a good Chet TV is not going to say here's the way to explain HSA is to a consumer. But what it can do is like, if you give it a very good prompt, it can write a really good article. So I think it's very similar to the coding thing where it's, can't understand the, the, like the content in mass and like how it interconnects and like how actually this person will learn.

But it can write a really good article and write a really good related article and write a really good related article that you can edit and pull together into like a pretty impressive web of content. So anyway, I'm super excited. That was an hour and a half. Now, if I'm writing something more technical, same, same piece, it requires maybe more research, more validation. I know HSA is already, I know the limits. know how to validate all the data in there.

Tyler King (25:00.338)
haha

Rick (25:14.957)
but if I was writing on something that I wasn't an expert on, would get very nervous publishing any of the content without a kind of rewriting it.

Tyler King (25:21.926)
Yeah. Yeah, I mean, that's awesome. And what I worry about with both what you're talking about and what I'm talking about with the vibe coding thing is that we're in this very narrow band where like you're providing enough extra value that maybe it's worth someone reading your blog post or using my app rather than making their own. I don't know, I find myself using Google Lesson Lesson just going directly to chat GPT, right? And like, I think there is still this band where

the super high effort human version is at the top and the full on just use ChatGPT yourself is at the bottom and there's this band in the middle where it's like someone else used ChatGPT for you. I do worry that that's not gonna last.

Rick (26:08.558)
What do mean, last?

Tyler King (26:09.956)
Like, like that no one's going to Google the things you're writing about. Basically.

Rick (26:15.478)
Maybe not. But like, I still think that having that content available that we've validated is content that we can feed our own models. We can feed into our product. It's ours. And so while maybe it's not gonna be the SEO value, content is still gonna be valuable, especially if you've sort of like, you own it and have validated the quality to pull into other workflows. I hope SEO

I hope SEO sticks around. hope there's some way for really good content to go to the top. man, like it just, it just doesn't seem like it's going to last if chat GPT can write a, you know, if chat GPT can eventually write a better article, know, art technical article than me, then you know, why would anyone try to try to go to Google to find my article when they could just ask chat GPT to write a better one.

Tyler King (27:06.204)
Yeah, exactly. I've been having a related conversation with Shelly, my wife, who, you know, she teaches stats and specifically like coding and R in college. And man, the kids, it's not their fault, but I feel bad for them. How do you learn anything when AI can just do it all for you? And especially coding, AI is so good at coding, and especially coding in R, which is like, it's just very basic.

really structured coding where it's not an art. Like you could argue that certain types of coding is an art, but being like, hey, do a linear regression on this or whatever. It's just like, what's the command? Here you go. And so as a result, none of her students are actually learning to code anymore, which like fine, maybe you don't need to anymore, but then she's going to them and saying like, well, you need to know something. And they're like, no, I don't, AI can do it for me. And the thing she has to go to them and say is,

If you don't need to know it, no one needs to hire you. Like, what value are you bringing to the world if all you're doing is plugging something into AI? And I'm not saying that's what you're doing right now, but I just think that's a thing a lot of us are gonna have to.

Rick (28:08.974)
Yeah.

Rick (28:18.988)
Yeah, one thing that I was thinking about when you were talking about her R thing, I like, just remember in computer science, like, I don't know about you when you were doing, did you do computer science or engineering? What you do?

Tyler King (28:30.418)
Yeah, I kind of started electrical engineering and shifted to computer science.

Rick (28:33.4)
So at Duke, the computer science major was pretty new. And it was at the time Facebook was being created, like there were a lot of cool internet technologies coming out. But the vast majority of the major, like what hooked me on computer science as a major was I loved the Java programming class. I loved like that experience. But then you start having to learn about all the registry stuff and how like bits and bytes, like all this other stuff. like, I don't care about any of this.

But like, feel like it's the kind of the same thing where it's like, you really like, maybe I did need to know that for whatever they were profession they were preparing me for, but like it says, it does seem like, you know, if you can produce the outcome with whatever tooling is available to you, do you really need to understand the bits and bytes of whatever it is? And I don't think the answer is you do, but there are some principles that you probably do need to understand in order to like.

Validate the outcome that your tool is using and maybe that's what you're talking about, which is like But like, you know with an HSA article, let me just go back to that. It's not just principles, right? Like there's actual legal statutes that have to be accurate and if you miss right like you're like committing financial it like miss malpractice like anyway

Tyler King (29:51.59)
Totally agree, yeah, you brought actual value there, but imagine you when you first graduated college being like, what, 95%, 98 % of what you did was done by AI and then you brought that extra 2%, which is important, not just important, it's critical, you have to have that. But if you don't know anything yet and you're trying to learn it, to be able to have something else do 95 % of the work, you never, how do you learn that 5 %?

that you need to know in that set. I've discussed, yeah, yeah, I've mentioned on this podcast before that I have this concern with customer service that as our customer service team gets more efficient with AI tools, it's gonna be harder and harder. The whole point right now of using an AI tool is it does a draft and then the human reviews it. But if someone starts working at a company and the AI draft is right 99 % of the time, how do they ever learn those edge cases so that they can catch the 1 %?

Rick (30:23.566)
How do you know what the 5 % is? Yeah.

Rick (30:50.51)
Another example of this same thing historically is whenever there's, I don't know if you've experienced this in less knowing CRM, but I experienced it at PeopleKeep and I experienced it at WinFall. We have internal tooling that we've built for employees to complete tasks. Those tools are always built to solve a specific problem that existed in the past. The business evolves, the process evolves. The new person who comes on is using that tool learns the tool, but doesn't understand the problem that was designed to solve and they miss the entire point.

Tyler King (31:15.388)
Yeah. Yeah, that's.

Rick (31:17.998)
I feel like that's the exact same thing that you're talking about here with AI.

Tyler King (31:21.424)
Yeah, and that's what you mentioned like when you major in computer science. I think it's still this way. You're still learning how to do binary algebra or whatever it's called and, you know, machine code and all this. And yeah, you don't need to use it. And you didn't even when we were studying it. But I do think like that's you a pure computer scientist, like an academic would say like, yeah, you need to know the foundation so that you can be on the cutting edge of what's next rather than just like.

10, 15 years ago, the coding version, we didn't have the LLM stuff, but you could be like, I'm just, I'm gonna install some pre-made open source library and install another one and they're gonna connect together. And it's like, okay, great, you can build a prototype, but you have no idea what's going on and you're not gonna be able to maintain it and all that. where are we going with this? Uh-huh, vibe coding, vibe writing, yep.

Rick (32:08.846)
So I was content vibing, I was content vibing, you were code vibing. No, vibe code, no, vibe contenting. I was vibe writing. Vibrating, okay, not content vibing, vibrating. Just changing this. Vibe writing.

Tyler King (32:23.442)
But yeah, that's cool that you got that content stuff. And you are in that really great position of being kind of like one of the world leading experts in a certain kind of niche part of the health.

Rick (32:34.764)
Yeah, I literally wrote all these articles before. I just don't want to like spend the time to write them again. and so chat TV is providing me massive leverage and honestly, these are better. I have one more thing related AI before we move on to topic. I did, decide to offer an AI internship to my, brother who is 17, no 18. and here's the, I was like, this might fit for thought. got plenty of time to figure this out, but I was trying to figure out what some good projects were for him to work on, but one.

Tyler King (32:38.598)
Yeah.

Tyler King (32:43.27)
Yeah, yeah.

Tyler King (32:47.666)
Mhm.

Rick (33:04.766)
really interesting idea. I don't know how you feel about this, but like one way you could create content is you could go like grab, go, go identify 10 competitors of X business, like go have an intern grab all of those, all of their content, pull it into Google docs and then have Jim and I basically assess those and rewrite them, different versions of them, under like whatever, you know,

company brand you want. Is that legal? Is that illegal? What I don't even know.

Tyler King (33:39.524)
Yeah, I don't think anyone does.

Rick (33:42.286)
So that's, but anyway, there's so many cool, like, I think internship projects on AI that a high school student could do or a college student could do. but you have to guide them in terms of like collecting and storing information to train a model. and then the model actually does the work. I think that's really interesting. something that we could talk about later, but anyway, I'm going to do this. I've got to figure out the right structure for it so that it requires, minimal time investment. for, you know,

Tyler King (33:58.93)
Yeah.

Rick (34:11.643)
the internship program has to be self-contained and give the instructions for the person to be successful.

Tyler King (34:18.556)
Yeah. Yeah, that's the is a legal question is definitely interesting. I think chat GPT maybe other ones do too. I think guarantees like they'll cover the legal costs if you get sued for copyright infringement. Now, I doubt that applies if you're like here is this copyrighted material. I want you to replicate it is different from please write me an article on topic XYZ. So I could imagine that getting you into a little more trouble.

Rick (34:45.326)
Yeah, that's interesting. Yeah, if it's got, if it's got copy, if it's copyrighted material, that's interesting.

Tyler King (34:49.874)
Yeah, but I mean, thing in LLM is great at is take this article and just change all the words and that I think that's like forget AI for a second. If you do that as a human, if you're like, I'm going to take this article and I'm just going to rewrite every single sentence using synonyms, I think that's no longer copyrighted, right?

Rick (35:07.918)
I don't know. That's a good question. But that would be the, if we answer that question, I would think it would answer the AI question.

Tyler King (35:12.902)
Yeah. The AI question, the bigger AI question that everyone else is talking about, which is not what you're asking is like the fact that chat GPT or Gemini or whatever learned, like built the model around all this content that out there that is copyrighted is that copyright infringement. And there's a whole debate about, well, if a human went and learned stuff by, know, if I, if I listened to music from a famous artist,

and that inspires me and I learn my taste in music from that and then I go write my own music and it sounds like it because that's what inspired me. That's how all art works, but if the AI does it, is it the same? That's a whole different discussion going on also.

Rick (35:53.154)
Yeah, yeah, interesting. Okay, so I'm gonna bring the AI internship stuff up again when I have it little bit more formed and get your feedback on the prompts. Because I need basically it to be something that someone can do with probably less than five minutes of guidance a week.

Tyler King (36:14.908)
Hmm, okay. Just to put a little bug in your ear, vibe coding is another thing. Like it could be build a tool for JD or something. Even if someone doesn't know how to code, you can literally go into chat GPT and say, I don't know how to code, tell me how to start this.

Rick (36:31.894)
boy, that scares me. You're really? Okay.

Tyler King (36:34.704)
Yeah, sure. I mean, I don't know how would go. I've read stories of people doing that successfully. A 17 year old maybe would have less like maturity in terms of how to approach that then. The people I've seen talking about it are non-technical founder types. definitely people have been getting, definitely, I wouldn't say don't let them touch my code base. But if it's like JD wants a button that's like send this email to this person, enter their email address and it sends the email. Like maybe that's possible.

Rick (37:02.444)
Yeah. Yeah. We'll come back. We'll come back. I want to, I want to, I want to spend more time on this than we have today. So, for, for, I just wanted to kind of shift to one topic real quick, benefit program. So that's what you're building right now. You talked about it earlier. I just wanted to brainstorm some pricing and packaging for you. You mentioned that you didn't, you weren't sure what, was going to be the benefit program types to me. Like I know that there are at least three, like generic types.

Tyler King (37:05.093)
Okay, okay.

Tyler King (37:14.438)
Yeah.

Rick (37:30.03)
Um, and I just wanted to call those out and I think it's kind of interesting, uninteresting pricing conversation, um, around how we might, uh, unlock some of the like, Hey, this is a no brainer to offer for free versus what you pay for. Um, so like right now we, we have these employees that are being added and they may or may not have an allowance aside to them. Um, w what I was thinking was maybe we offer a free employer program where they can still take advantage of our non snake oil.

value added employee adding plus a benefit program for like purely health insurance concierge, which does not allow any contribution. and so like, like it, and the free package is the employer can add and remove employees. They can, they get one benefit program for the health insurance concierge, which is always available for every company, whether they add people to it or not. it's like this, it's like the starting benefit program. Would you like to add people to the free health insurance concierge? it's, here.

add people, add your part-time people, add your full-time people. There's no cost to add people to this. or maybe it's included in a base fee or something like that. I haven't really thought about it, but it's not PPM per employee per month pricing. but then, you know, there's like, would you like to set up another benefit program? And then there's basically two types of programs within that. one is a, know, you want to reimburse someone, for an expense, meaning you want to adjudicate that they are having some sort of expense and cover that.

based on some rules or you want to just provide the money for a specific purpose. which is a stipend piece. And so like those to me are like the three types to begin with that are massively flexible. and, you know, the first is just like standard out of the box and you only need one of them because it's purely the health insurance concierge without any allowance. What do think about that?

Tyler King (39:23.258)
Yeah, the thing I'm least compelled by is the stipend thing, which is kind of what we already have.

We don't pay the employee the money. The employer does that through gusto or whatever payroll. So I'm not, that's the, to be clear, I'm not saying we're selling snake oil or anything, because JD's service is very valuable, but like, we're just displaying the allowance they get on a webpage. I'm not really sure where the value is there.

Rick (39:35.608)
Payroll, yeah.

Rick (39:49.708)
Yeah. So, so where, where the long-term value is not where the value is today is around help. Like the biggest challenge with offering a stipend to an employee is, is not actually giving them the money. It's not actually deciding to give them the money. It's not actually adding them. like in the software it's communicating, Hey, this is the benefit. Here's why it's valuable. and doing that at different stages of the journey, and making it feel like a benefit, versus.

some just more money on my paycheck. And that is like the concierge piece. So it's kind of going back to the concierge piece, but plus a stipend. Now, whether someone should pay PPM for that, what the pricing is a whole nother discussion, but I do think like, it's different. It's a distinct different type of benefit program than the health insurance concierge without a stipend. Because what I just want to make clear is like, from the employer's perspective, when they add someone to a health insurance concierge program, there really is no cost. It's

pure upside for them. When they add someone to a health insurance concierge, to a stipend program, even if we were free, there's a huge cost to them. And so the stipend itself, so making sure that those are two separate things, distinct things, even if there's no significant pricing difference between the two, although I would think that we long-term will offer the health insurance concierge for some very low price compared to like,

Tyler King (40:58.514)
the stipend itself costs money.

Rick (41:18.604)
whatever else we do with benefit programs. I think that that is something that would unlock potentially a ha moment for an employer going, that's why I should set up an employer account with you guys for free and add all my employees.

Tyler King (41:38.364)
Sorry, you're saying that this like positioning or it's not positioning, but whatever, this like way of packaging these three different types of benefits programs makes it more clear why the thing we're already offering is valuable.

Rick (41:43.01)
More packaging.

Rick (41:52.334)
Why why you should right now? One of the biggest challenges we have So a lot of what we're right now is for customers who are paying us for giving people money in some in different ways There is a whole pool of people out there that we cannot have not successfully convinced That that they should set up a free leg up employer account and add all their employees part-time full-time That is the packaging that I think is potentially

app available through the benefit program as a base program. I don't know if it's the right play or not. just brought it up as a brainstorming.

Tyler King (42:29.5)
Yeah, I don't have good intuition on this. This has been the most frustrating thing about LegUp Health from the beginning, is that this is an absolute no-brainer. There are a lot of small businesses out there that don't offer health insurance or any health benefits at all. Why would you not do this free thing? Which for people listening, I think some of the language might be vague. Basically it's saying, if you add your employee to this software, JD will...

help them get health insurance like any agent would. They could go to JD individually and do it. It's just like, you don't have to do, know, every employee does not have to fend for themselves now. They all have like this one contact person to go through.

Rick (43:11.502)
And more importantly, like you are basically providing them a resource versus them. If you go online and search for health insurance, the likeliness that you will get a helpful agent is very, very low. You're going to most likely get spammed by 20 agents because you're going to accidentally put your name and phone number into the wrong thing. So there's actually like, it's more just like, Hey, like connecting the employee with a resource.

Tyler King (43:32.528)
Yeah, we vetted this person. They're legit. And that mirrors my experience quite a bit because for a long time, Lesson on Serum did the, don't have health benefits, go get your own individual plan. Like, I'm a big believer that insurance should not be tied to employment. a bunch of gradually over time, employees kind of push back and were like, no, we want a group health insurance plan. And the reason wasn't that they wanted better insurance and it wasn't the cost, it was the hassle. They were like,

Every year I have to like log in to healthcare.gov or whatever and they offer me these, I don't know what these things mean. Can you just pick a plan for me? That's basically what the, so the reason they wanted group insurance was for the simplicity and we're offering them that simplicity but not the paying some of the cost and so on.

Rick (44:23.018)
Exactly. Exactly. So like, I think if, if we can figure out how through the software give JD the ability to demo and make it clear, Hey, this is what you get for free. This is what you get out the, or the base, like the base package. might be 50 bucks a month. I don't know. you can add as many people as you want to do this. Here's the value that we just talked about, blah, blah. by the way, if you want to upgrade, I don't know what that is. We have these other benefit programs that get into like actually giving people money.

on various rules, for example, you could do stip, stipend programs or reimbursement programs. and you know, here are the two packages for that. I don't know. I'm not saying that we should do that. I'm just, I'm, I'm more just saying like making the health insurance concierge offering more tangible, without the stipend and its own thing, I think could be an unlock.

Tyler King (45:02.268)
Yeah

Tyler King (45:14.652)
Yeah, I'm not sure. I'd like to see JD go and try that. He could do that right now, presumably. We've struggled so much. I haven't talked to anyone ever about this. He and you have tried to convince people, hey, this free thing. I don't know that having that this packaging changes much in my opinion, but I'd definitely be down to have him.

Rick (45:37.902)
The main thing that we fall down on is we don't have communicate like we are, we are just giving them what you've said, which is like, it started as a spreadsheet, which is now basically a credit app of adding and removing employees where we, where we fail and we don't provide like tangible value is on the communication of the value to the employees. So basically providing the employee like health insurance concierge out of the box could be here's a blurb for your employee handbook. Here's a blurb for a new hire.

Here is a one pager that is custom to your thing. All things that JD could do, but require templates and that sort of thing.

Tyler King (46:10.064)
Yeah.

Tyler King (46:14.354)
That feels separate to me from the rest of this packaging thing of, and then we've also got this stipend thing and we've also got this reimbursement thing. Yeah, I buy that trying to make, trying to build more communication around the whole thing, not just the free version of it makes sense. Yeah. And JD.

Rick (46:34.166)
A benefit program. So, so what, what, what, what I'm going, where I'm, I want to make clear is like not the whole thing, is in the platform. I'm talking about like a distinct benefit program has distinct communications based on who is eligible and participating in that. And so I guess what I'm trying to say is like, we could just make the benefit programs about giving people money and communicating that. But I think there is value in taking the same approach to the benefit program of just the health insurance concierge, which is just more about like.

Making sure we have the flexibility to do that more so than, you know, and, and also like making it easy for JD to demonstrate, Hey, like the, know, you can set up a benefit program with no contribution. can set up a benefit contribution, a benefit program with a stipend contribution, or you can, set up a benefit program with reimbursements. and you know, here's how, how that works. And so for example, like you could, let's say you're, you have a benefit program where you are.

paying for people's cell phones. But you're not going to like track that through the system. You could just say, Hey, you know, I'm giving you a company sponsored cell phone and you could set the rules for that program. I use this as a software to communicate, communicate the value of the program and that they're eligible for the program through the leg up benefit software. But you're not going to give them a contribution through the software, for example. Anyway, I think I'm, I think I've made the point. I just want to make sure it's clear.

Tyler King (47:52.487)
Yeah.

Tyler King (47:56.946)
Okay. Yeah, if I were a customer, I'm still a little like, that was a lot of words and I'm not sure I fully get it, but...

Rick (48:06.754)
That's, and I think that is part of what JD in the market is, is challenged with is like, think this is much simpler, like visually just described. And if you use their interface more so than like him talking about it. And that's part, that's part of what I want to want to want to test is, like if we gave it, we said, Hey, here's a, here's an account. You were giving you a free benefit program called health insurance, the health insurance concierge program. You're not using it. Would you like to use it? I think there people, I think any company that

Is already buying like I'm thinking of all the business owner consumers who are buying health insurance through us right now that have 10 other employees that they're not offering health insurance to. was like, yeah, I didn't even know this existed. Like I'd love to add people.

Tyler King (48:40.956)
Yeah.

Tyler King (48:45.13)
I don't get what's stopping us from just doing that now. Like I feel like we don't need anything changed to just reach out to them and be like, sorry, my intuition this whole time has been, what leg up benefits is right now is you can do this, we'll be your health insurance agent, what you're calling the health insurance concierge, that's free. Or you can pay us, I think the pricing differs by customer, but something like $20 per employee per month to get the stipend thing.

Which is the exact same thing except we just tell you you have a stipend. I've always been of the opinion, who in the fuck would pay $20 a month to tell someone that they're paying them $200 a month or whatever the number is in stipend?

Rick (49:27.832)
That's why I think that's why I'm trying to make it clear that's not where the value of the benefit program is. Yeah.

Tyler King (49:33.106)
Well, yeah, but hang on a sec. I've been of the opinion that that made no sense as a value prop and the free thing of you're getting like 99 % of the value of the paid thing, which is JD's service, that's a no brainer. And yet, and yet he's having an easier time getting people to pay money for this, what I view as a worthless add on. And it's incredibly difficult to get anyone to use the actual super high value free thing. I don't get it.

Rick (49:44.192)
Yeah, for free. That's the question. Yeah.

Rick (50:01.644)
Yeah, I mean...

Tyler King (50:02.588)
So my intuition, don't trust my intuition here for that reason.

Rick (50:05.772)
Yeah, I know. I think, I think you're right. part of what I'm, I think what I'm hoping to do is like use this benefit program to further, validate what the issue is. so like, I, I know that when JD is given a tool that he can put in front of people, it's going to elicit more learning that he's going to be able to get back to us. and I have, I have a sneaking suspicion.

Tyler King (50:16.38)
Hmm, yeah.

Rick (50:33.304)
that if we think through the packaging of this benefit program, helping him explain what we do, and we do it flexibly enough, we're gonna have an unlock. I don't know what the unlock is, but it could be free health insurance concierge and free stipends and then paid reimbursement. I don't actually know.

Tyler King (50:52.71)
Yeah. So to make sure I understand what you're saying, I'm building this thing that you and JD both asked for. And I'm asking the question, well, what are people actually going to use it for? And you're saying, we don't know what they're going to use it for. Build it. JD can go demo it to people and we'll figure out what they use it for. And then we can kind of refine it.

Rick (51:09.774)
We know that we have customers using it that will use it for the following things. A health insurance concierge, no stipend, which is one type of benefit program. People are paying us for that today. They're not paying $20 per employee for it, unless they are sub 10 employees. So we generally have a floor of a hundred to $200 per month for this product. Some of them are using it for stipends and some of them are going to use it for reimbursements. that's, that is what people will use it for today. What I'm less interested in,

I'm certainly interested in how they actually ended up using it, but I'm more interested in how JD uses this to demonstrate the value to someone who he is not selling group health insurance to, and not setting up on an HRA. think like where we're winning is group health insurance. We're winning on reimbursement of HRA, but I feel like we lose a lot more than we win on the stipend and. Health insurance concierge, like I'm not offering any benefits situation.

Tyler King (52:08.242)
Okay, well, I'll build it. I'll be curious to see what happens. I just feel like, yeah, I feel like fundamentally the software won't do anything new that it doesn't do now. Like it already is a benefits program. It's just one instead of more than one. That's what I'm building, is the ability to have more than one within a single platform.

Rick (52:11.406)
I can't tell.

Rick (52:30.56)
And the ability for JD to build, add-ons around this real thing versus, like, think the communication is the key piece here. It's not going to be built. It's, he is not doing it now. and I've watched customers ask for it and we don't necessarily deliver the right way on it. And so it's like, yeah. Yeah.

Tyler King (52:40.828)
But he could do that now if he wanted. Nothing's stopping him from doing that.

Tyler King (52:51.418)
Nothing I'm building has to do with communication,

Rick (52:54.85)
I'm working on that with him, but you do have like a link for us to link in the benefit program, like plan documents, which are part of the base of the communication. And so I think this is going to provide more structure for us to innovate within. And then we'll learn. But man, I didn't mean to demotivate on this. That was not my... Yeah.

Tyler King (53:15.086)
no, I'm not demotivated. I totally agree. The question is what can JD sell?

Rick (53:24.428)
He can sell right now. He's selling group health insurance. is selling consumer health insurance. Okay. It's great. Okay. Okay.

Tyler King (53:27.986)
Sorry, I'm not asking you that question. I'm saying it's a really interesting, this is not where most companies end up. There is tremendous value in leg up health. When someone uses us, the service that they get from JD is tremendous and free, but it's hard to sell. we have this product, normally the product is the thing providing the value. And here the product is basically,

it's packaging and marketing for the underlying value which is already being provided. It's just a different thing that I'm not used to. And so it's like, I'm building a marketing tool for JD rather than I'm used to building a CRM where it's like, it's gotta make you more productive and you've gotta live in it all day and every, this took six clicks before, now it only takes five clicks and none of that shit matters here.

Rick (54:01.239)
Correct.

Yes, yes.

Rick (54:13.76)
Yeah.

Rick (54:19.404)
No, this is about simplifying in people's head what they are doing for their employees and making it easier for them to explain it to employees, explain it to themselves, and then explain it to employees and, as a real thing versus money on their paycheck. That's all it is.

Tyler King (54:35.858)
Okay, well, I'm sure this will be an ongoing discussion. I'm definitely, you know, I'll build it, we'll see what happens, and I'm curious to see. Sorry, sorry, that sounds like, what I mean is I don't know what to expect. I'm not saying I'm, I would be pushing back if I thought this was a bad thing to build.

Rick (54:51.745)
Okay, good, good.

Tyler King (54:53.842)
yeah, cool. No, that's all good. I'm not, I don't have anything to talk about, so.

Rick (54:55.832)
Sorry, I dominated today on the League of Hell stuff.

The, so, I want to spend more time around the packaging things. I think we actually got to some good places there. I, I actually have a clear idea of what I'm trying to say. And then I think you also had kind of picked up on what I was saying. And I've, think we were calling it a marketing tool for JD now versus what we, whatever we were calling it, which actually I think is the right way to look at this next version. So I'm going to take that as an action item to think through like, okay, wait,

What are we actually solving for? Why do I want to, yeah.

Tyler King (55:33.764)
Mm-hmm. Yeah, like here's one just constraint you can put on yourself there is like, is the, you you have this pipeline for JD where it's like, you know, email, phone call, whatever, and there's a demo call at some point. What is he actually showing the person? Anything that doesn't show up in that call shouldn't be built.

Rick (55:58.968)
That's actually really, that's a, my hypothesis, let me just state this clearly. My hypothesis is that the reason that our free service for health insurance concierge or very cheap service for health insurance concierge for companies that do not have for health insurance is not flying off the shelf is we are unable to successfully demonstrate the value. That is my hypothesis. And I believe that a, think that, I believe that having a tool

that we building some demo, some functionality of the tool that is very basic, not necessarily the advanced version of this to demonstrate this was, going to be the unlock. I agree that there are other ways to do this through decks and sorts of things. But, that has not, we've tried that. So, or at least I think we have, I shouldn't say we have tried it cause I don't actually know.

Tyler King (56:51.516)
Yeah, okay, cool. Well, we probably gotta call it here. I'm sure this will be TB continued. I will. See ya.

Rick (56:55.48)
Thank you for your time. Go take a nap.

All right, see you later.

Vibe coding
Broadcast by