Subscribe
Share
Share
Embed
Aaron VanStone built ScriptSafe, an AI-powered pharmacy-accreditation platform challenging an industry with only two gatekeepers, coding the entire stack himself with zero prior experience using Claude Code. He breaks down the workflow, the cheap starter stack, and the hard-won lessons behind it.
"The classic thing they teach you in every coding class is that the more preparation you do beforehand, the better the end output's gonna be. It's no different with these models."
Most companies are buying AI tools. Almost none of them have the data, processes, or systems in place for those tools to deliver.
The CRM is a mess. The processes aren't written down. The tribal knowledge is locked in three people's heads. Then the AI gets layered on top, and nobody understands why it doesn't work.
AI-Ready or Not is the show about fixing the foundation first.
Host Travis Scott talks with founders, operators, authors, and practitioners about how they're actually using AI. What's working. What's not. What they had to do to make it work.
New episodes every two weeks.
AI-Ready or Not? - Aaron VanStone - final
===
Travis: ~Uh, and even though it's recording, I, I can cut a lot of stuff, so, um,~
Aaron VanStone: ~Sure~
Travis: ~so yeah, just, um, gonna have just a casual conversation, kinda like what we had at, at Starbucks and,~
Aaron VanStone: ~Yeah~
Travis: ~just want to kinda share your story and, and how, um, how you're using AI to be disruptive in an industry that, had, had a large moat for a very long time. So,~
Aaron VanStone: ~Yep. Yep. Yep~
Travis: ~that's a good, a, a good story. So, um, so yeah, and then, um, I can add kind of more about, about you, but, um, kinda in the intro. But if you wanna just kinda, uh, just kinda say what, what you do, where you work and, and things like that, then I'll have it in the transcript and can pull it together,~[00:00:00]
Cold open
---
Aaron VanStone: I feel like it's, it's almost like divine intervention that these tools came available right at the time that we needed these because what we built, that wasn't possible a year ago.
Welcome to AI Ready or Not. I'm Travis Scott, founder of Rainier RevOps, and this is the show where founders and revenue leaders figure out what it actually takes to make AI work in their business. Let's do this.
Meet Aaron & ScriptSafe
---
Aaron VanStone: So my name is Aaron Van Stone, and, ~uh,~ I'm building a company, ~uh,~ named ScriptSafe. ~Um, ~so it's in the pharmacy space, and currently, ~uh,~ for a pharmacy to accept online payments, ~uh,~ or, ~uh,~ place an ad online, like via Google, you need to be accredited by a third-party accreditation company.
And for, ~uh,~ it seems like since the beginning of time, there's only two of those, ~uh,~ in existence. ~Um, ~and because of that, [00:01:00] there's a lot more pharmacies that need this, and there's been a, a significant backlog because of that. ~Uh, ~and so because of really starting this, ~uh,~ unbeknownst that these AI tools would be available to us, there's no way that we'd be able to build it without it.
~Uh, ~but we've been able to build a, ~um,~ pharmacy accreditation company that not only looks over all of their current licenses, but monitors their websites each night so that, ~um,~ we can confirm that they're not selling any illegal, ~um,~ illegal drugs online. So, ~uh,~ we hope to go live with this this summer.
We're waiting on, ~uh,~ literally a incredibly large credit card processor and Google to, to give us the approval, and we're deep in those talks to get that done.
Travis: ~Awesome. And, um, cool. So then I can use that and kind of create an intro and, um, I like to start the podcast kind of, uh, when I, when it's, when someone's listening to it, to jump kind of right into the~
Aaron VanStone: ~Yeah~
Travis: ~and then I'll kind of cut, add some context, and then kind of keep going, um,~
Aaron VanStone: ~Yeah, that makes sense~
Travis: ~with it, uh, just to kind of draw them in.~
~Um, ~and so, ~um,~ cool. And so, ~uh,~ so yeah, before you started ScriptSafe, ~uh,~ what were you doing before, before that?
Aaron VanStone: So f--
From payments consulting to a new problem
---
Aaron VanStone: I've been in the payment space, ~uh,~ almost my entire career. So I started a [00:02:00] consulting firm called The Processing Brokerage. ~Uh, ~and what it does is it basically looks at a merchant's credit card processing statements and sees if there's ways to reduce fees. And then if we're able to do so, we get a portion of those savings.
And so it's been... I started that in two thousand and nine, and, and really it's been, ~uh,~ what I've, what I've used as my job really, ~uh,~ for, for a very long, over fifteen years now. ~Um, ~but when COVID happened, ~uh,~ I realized that even though we were just fine, that I was becoming pretty dependent on a couple big clients, and if I were to lose those clients or some disruption like COVID where I didn't lose them, but I could've, ~uh,~ I would've been in tough shape.
~Uh, ~and so it was at that point where I started looking for, ~um,~ maybe some other opportunities that may be allowed to kind of diversify a little bit and expand the consulting within the payment space of looking for other opportunities within [00:03:00] payments. ~Um, ~and a couple years later, this opportunity with, ~uh, uh,~ the Alliance for Pharmacy of Compounding.
They are a, a trade organization, ~um,~ advocacy group for compounding pharmacies, and they wanted to build a credit card processing rate program. So I started working with them on this, and then that's what kinda led us down to figuring out this whole, ~uh,~ what ended up being ScriptSafe, just because we were seeing so many, ~uh,~ just kind of, ~uh,~ bottlenecks within the industry that just, that just shouldn't be there.
~Uh, ~and so then we started to try to figure out how we could solve it.
Travis: Yeah. Yeah, that, that was gonna be my next question, is kinda
Divine intervention: finding AI at the right moment
---
Travis: how, how did the idea come up and, and, ~uh,~ you, you ran into the problem and now you're trying to solve it. a- at what point, you know, as you were exploring that, was it immediately, ~um,~ were you immediately kinda cognizant of AI could help solve this?
Or were you kinda [00:04:00] exploring, doing research, and then you came to the realization, "Hey, I think we can do this with AI"?
Aaron VanStone: So let's see if I can-- When, when you and I met in Starbucks, I think I fumbled through this explanation. It would-- might've taken me 20 minutes 'cause i- there's, there's so many steps. So I'm gonna try to be as condensed as possible. So this time last year, we were going to Visa, ~uh,~ 'cause Visa controls these rules, ~um,~ to try to get an accommodation.
And so Visa changed. ~Uh, ~we were able to get that accommodation through Visa. And so then, ~uh,~ it became clear that starting our own accreditation company was the way to go. So to answer your question about knowing AI, I feel like it's, it's almost like divine intervention that these tools came available right at the time that we needed these because what we built, that wasn't possible a year ago.
~Uh, ~and, and so it was literally right when we were needing this is really when they were taking off. So I [00:05:00] started at the end of January. I needed some sort of, you know, website that I could capture this information for the accreditation, ~uh,~ and then also have a CRM in the back, but it was unique. So you couldn't use something off the shelf.
You needed something built. And, and going to a programmer was obviously going to cost a lot of money. And, ~um,~ so then that's when I was just like, "Well, I've been reading a lot about this vibe coding that everybody was talking about back in December and January. Let's try this out." And then, you know, you, you have no idea what you're doing, or at least I didn't.
~Uh, ~but then you start seeing like, wow, this actually might be able to work. ~Um, ~you know, and you're clumsy at first, but as time goes on, you become pretty savvy on how these tools work and, and what they're capable of. And, and I'm blown away each day, ~uh,~ by, by what they're doing. ~Uh, ~I mean, as-- it's case in point, ~um,~ what we built just this week with [00:06:00] the new version of Claude is just incredible.
So
Travis: Hmm.
Aaron VanStone: yeah.
Travis: Yeah, and
Fable 5 vs. Opus — you feel the upgrade
---
Travis: that is the, ~uh,~ the, the Fable five, right?
Aaron VanStone: Mm-hmm.
Travis: That version?
Aaron VanStone: Yeah. Yeah. Yeah. A-and what's, and what's funny is, is I don't know if you've noticed this, but, ~um,~ 'cause I've been in a mode for like the last four or five months where I've, I've got my laptop open coding, you know, going through Claude Code and also using Codex and, and a bunch of other different models now at this point.
~Um, ~but because you use them so much, you really do notice when there is an upgrade because the way they respond, the way that they provide the results that you're asking for, ~uh,~ you, you see it pretty quickly. And, and this Fable 5, it just is... It was always possible before, but now it's just, it's polished and it's super easy.
~Uh, ~and I'm almost through all my data limits and I still have four or five more days left, so I, [00:07:00] my, my, my coding might be a little less this weekend.
Travis: Yeah, yeah. Are you using Pro or the Max of
Aaron VanStone: using the Macs. Yeah.
Travis: still,
Aaron VanStone: Yeah. Still, yep, yep. Yeah. Yeah.
Travis: 'cause then after, after you bump that, against that, then it's, it's kind of a, a pay per, ~um,~ token type of thing, right?
Aaron VanStone: ~Uh, ~that's exactly right. And, and I saw that Claude is going to allow the subscriptions, which is what we're talking about, to use Fable until I think like June 20th, and then you have to pay for it moving forward. So it's, ~uh,~ not to speak ill of anyone, but it almost kind of feels like a drug dealer kind of giving you a, a special deal to get you addicted and then, and then, ~uh,~ and, and, and then they're gonna try to keep you, keep you paying afterwards.
Travis: Yeah, e-exactly. Yeah,
Aaron VanStone: Yeah.
Travis: think everyone kind of wondered how are they gonna continue to, to monetize things
Aaron VanStone: Right
Travis: right? Because, I mean, [00:08:00] Max might seem like a lot, but what you get for it, it's extremely cheap. ~Um, ~compared to what you would pay someone to do this stuff that you're doing, it,
Aaron VanStone: Yeah.
Travis: very cheap, so
Aaron VanStone: Very cheap. Very cheap. And, ~uh,~ I mean, ca- ~uh, ~this yesterday built a website basically off of a flyer, ~um,~ and then also, ~uh,~ built a portal where you can upload statements and different things like that. And I mean, and this was while I was watching, ~uh,~ the first World Cup match, so it probably took me about 90 minutes to do, and I really wasn't grinding that hard on it, and it looks, like, incredibly polished.
Like, it's, it's pretty, it's pretty jaw-dropping what, ~uh,~ what these models are doing. And it... And I still shake my head of where things have come from just from January, not even talking last year or the last five years, just from January. And it seemed like every month they were-- it was monumental increases that they were making, and we're at that [00:09:00] next monumental increase
Travis: Yeah, and I mean, I have, I, I've used Fable 5 a little bit. ~Um, ~I haven't been in code a lot, although I am today, ~uh,~ creating an app, ~uh, uh,~ an Instagram for analytics o-
Aaron VanStone: Yeah
Travis: so I can start to analyze some things and figure out how can I grow this better.
Aaron VanStone: Right
Travis: am trying to build for my nonprofit because, you know, are tight and
Aaron VanStone: Mm-hmm.
Travis: hire anybody.
My time is, is really tight and trying to figure out, and I hear people talk about, "Yeah, I have, ~uh,~ an that, you know, helps with content creation and social media and, and
Aaron VanStone: Right
Travis: interview and breaking it into clips." And I'm using Opus Clip for that now, but would love to see what I could build on my own and
Aaron VanStone: Right
Travis: to leverage...
I mean, I'm still creating all the content, but just some help, right? ~Um,~
Aaron VanStone: Right
Travis: an intern to come help with some of this stuff. ~Uh, ~they're not gonna shoot, [00:10:00] you know, they're not gonna talk into the camera about stuff they don't know about. I am,
Aaron VanStone: Right. Right
Travis: the help. So trying to build that now, ~uh,~ it kind of, it kind of sucked because, ~um,~ I was right in the middle of building it and I needed to get some tokens from Meta, ~uh,~
Aaron VanStone: Yeah
Travis: Facebook went down today. And so I was just like, "Did I just crash?" Like, a- and there's just a blank screen. I'm like, like, "This, this isn't supposed to happen. What's going on?" Like, ~um,~ and then I went to Claude Code and I'm like, "I think Facebook's down." He said, "Let me check." And said, "Yep, it's down."
Aaron VanStone: Yep, sure enough
Travis: right." ~uh, ~but yeah, so I haven't used Fable 5 much since it came out.
I also did notice that, you know, it said after June whatever they're gonna start charging. I'm like, "Well, do I really wanna like addicted to it and,
Aaron VanStone: Right.
Travis: like, oh man, like Opus wasn't good enough when Opus was just fine for
Aaron VanStone: Was just fine. Was just fine. And, and literally I was singing its praises this time last week. I'm like, "Man, it's so much better than the previous version, and, and this is just [00:11:00] working so much better." And yet now it- that's just old news.
Travis: Yeah, exactly. Exactly. And so since you have been working in it a lot, especially over the last week, ~um,~ what are s- some of the biggest differences you've found, ~um,~ between Opus and, and Fable?
Aaron VanStone: ~Um, ~y-you know, the, the biggest deal is that it's, it's, it's just so much more polished. ~Uh, ~so like, like the example that I always kind of like use is, so all the work that I've been doing,
The 'building a house' analogy for model progress
---
Aaron VanStone: I've been building a house basically, right? And so the earlier models, you would say, "Okay, I wanna build this house." And then they would say, "It's done."
And then I'd go, "Well, where's the kitchen?" "Oh, well, I built the kitchen, but the kitchen's on the other side of the street. You can see it. It's there, but like it's not in the house," right? So then, you know, as time goes on, now everything was in the house, but maybe it was more of like a tiny home. So everything was there and it was pretty basic, but it, but it-- you did have the house.
And so then, you know, as you move along, [00:12:00] now it-- this, the fable, it... carrying on with that, ~um,~ you know, analogy, it's, it's almost like it's asking you, "Well, do you wanna live in a Mediterranean villa? And if so, I would pick this marble for like this flooring because this would match ni-" Like the level of detail and polish, so is, is pretty incredible.
Now with all of these, and I, and I think you know this quite well, and I'm sure the majority of the people listening to this would know, ~um,~ the more detail that you can give in your prompt and really go through so much of the logic that you know that it's gonna be kind of requiring or asking for is it just, it's, it's, it's a, it's a game changer as far as the output that you will get on the other end.
And that's probably like, if at some p- you know, I, I don't know along this conversation of if there's any tips or tricks that I've learned along the way. ~Um, ~what's funny is, is that
Prep beats prompts: plan before you build
---
Aaron VanStone: the classic thing that they tell, that [00:13:00] they teach you in every coding class is that the more preparation you do beforehand, the better the end output's gonna be.
And it's still, it's no different with these models. If you are very descriptive and exact on what exactly you want, then, then you're gonna get a really good product. ~Um, ~and this, this new model just does a really good job of kind of filling in the gaps that you may not have, ~um, um,~ put in your description.
Travis: Yeah. And, and, ~um,~ I've noticed that too. And, and I used to write these big, long prompts and still, you know, would work through it, and it'd be, you know, 80%, fine-tune. It'd take me forever. so now, and I think we talked about this at Starbucks, in my instructions within Claude, I've instructed it to basically reverse prompt everything I put in
Aaron VanStone: Yep
Travis: always ask me questions to get more context. ~Um, ~and I think part of the challenge is, ~um,~ you know, when you're building something, you're building it based on your experience [00:14:00] in an area, and you've
Aaron VanStone: Yes.
Travis: doing that
Aaron VanStone: Yep
Travis: a while. And it's kind of the same concept of it's hard for very experienced people to teach beginners because they forget what they had to learn in the beginning, right?
Aaron VanStone: right. Yeah
Travis: start from ground zero.
Aaron VanStone: Right
Travis: like it's the same with prompts of, I know what I need, it's in my head, but how do I describe it to, to, ~uh,~ technology that might not know or have that experience or that experience that everybody as individuals has, right? ~Um, ~and so I think that's the challenge.
And so I, I, ~um,~ just built in the reverse prompting right
Aaron VanStone: Yeah
Travis: ~Um, ~and it... Sometimes it's annoying, but I know it's saved me a, a lot of time of trying to think about everything and then just getting it down to finally being what I wanted originally, right? And so, so that's, that's helped a bunch. ~Um,~
Aaron VanStone: Yeah. A-and I've, I've, [00:15:00] I've followed that as well. ~Um, ~and I think one of the other things that has really helped me is when I start to do, like for instance, ~um,~ I created a new website for ScriptSafe. ~Um, ~one of the things that I did before anything was built was
Running Claude and Codex against each other
---
Aaron VanStone: I asked Claude and Codex, so I have them working together, and I have to make sure that if Claude proposes something, Codex agrees to it.
So sometimes the two of them are kind of battling in the, ~uh,~ chat script, right? ~Um, ~but to ask them to make a working document of what this website is going to be, and then ask me every question possible of what, of like everything that it needs to be. And, and I would give it, you know, kind of details of this is what I'm looking for, what am I missing, kind of the reverse prompting stuff.
But that process probably took maybe 90 minutes to two hours of going through and answering and like reading and then waiting for those two to go through. ~Um, ~but [00:16:00] what ended up happening was, is that then you have this incredible document that you can say, "Okay, now build this." So then I'm not having to put a bunch of small prompts in.
You're basically just working off of everything that was already published in this document. And then, you know, you're saying, "Okay, today we're working on chapter one. Build all that." And then tomorrow, chapter two, and, and it-- and you get a cleaner product and, you know, and I guess the thought is, is that all of the wiring and plumbing is gonna be a lot more accurate because it knows what it's building.
Travis: Mm, exactly. Exactly. And I think, that's a good point up I think people can be worried like, "Hey, anybody can build anything with AI, so now, like, that bar has been lowered." But I think there's still gonna be [00:17:00] factors that allow some people to be successful and others not, and I think part of it is just that time you put into it ahead of time.
I think so many people just think it's an easy button, just type it out, and it goes and builds it all. ~Um, ~but there's a lot of work that goes into the planning in the very beginning of
Aaron VanStone: Yes
Travis: you even start, right? And I think that's gonna be the differentiator for people who are successful in using AI to build things and those who, who are not and just create slop, right?
And I think we're gonna all be attuned those little nuances to be able to pick up this is slop and this is solid, right? And
Aaron VanStone: And, and we've seen that, right. Because we, we've seen that just on social media with images and videos and things like that, where before like, "Oh, I don't know if this is real or not." And now already you go through like, "Yeah, that's, you know, AI." Like funny, great, whatever. But then, you know, some are good and some are like really poorly done.
And so you [00:18:00] already see that and, and now we're about to hit this next phase of software is gonna be the same thing. And is it... Yeah, i- is it the AI slot version of software? And I think that there's going to be-- it's, it's gonna be an interesting time over these next two years of companies going all in on, ~um,~ on AI and building systems that they really depend on, and now let's see if, ~uh,~ let's see how well it works
Travis: Yeah. Yeah. A-and, and if it does work well, then, and I, I'd mentioned this to my wife a few weeks ago,
Vendor lock-in and the coming software shakeout
---
Travis: now they're gonna be at the mercy of of keeping the prices reasonable because they can... Now you're in, your whole business is now pinned on this thing,
Aaron VanStone: Based on it. Yep. They, they have you hostage
Travis: 50% increase in prices.
What are you gonna do?" Right? Like, so, and switching can be done, but it's gonna be very difficult once you've built time, right? And the longer you've built in it, the harder it's gonna be to, to move to something [00:19:00] else, I think.
Aaron VanStone: Almost like when you were beholden to a certain software that your business was built on.
Travis: Mm-hmm. Yep, exactly.
Aaron VanStone: it's the same thing. Now it's just, now it's just another player.
Travis: Yeah, exactly. And I
Aaron VanStone: Yeah.
Travis: I think the next generation of AI, ~uh,~ where it'll really get shaken up is when people realize to cross that moat and, and break down those barriers, make switching easier, right? Make it
Aaron VanStone: Yep
Travis: elastic, ~um,~ and put that pressure, ~uh,~ that decrease that pricing pressure, ~uh,~ strength that these companies w-will have eventually. So I think that's
Aaron VanStone: I, yeah, I, I, I fully agree. And, and what I'm curious, and actually this is a question for you because this is something that I struggle with, is how do you find, how do you find other vendors?
Travis: 嗯。
Aaron VanStone: Because like, y- you know, to me almost like YouTube might be one of the [00:20:00] areas. I mean, maybe Reddit, like things like that, because I can't go on Claude and say, "What's a better version of you?"
Well,
Travis: Yeah, exactly.
Aaron VanStone: it's, it's not gonna respond kindly. You don't do that. ~Uh, ~and, and you can't really put in a Google search because that just gets back just the same typical responses. ~Um, ~so, so I think that's gonna be, I think that's gonna be a challenge for some of these vendors is how do I find out, how do I find where there's another vendor out there that can, you know, some of these API calls that I'm doing for ScriptSafe that are, that are gonna be becoming expensive with Anthropic, is there another vendor that I can use instead that gives me the same results?
You know, and how do I find that?
Travis: Yeah, e-exactly. Exactly. ~Um, ~yeah, I think that is gonna be challenging. I think also, ~um,~ similar to that is so many companies are building their own things and I, you know, I've-
The 'vibe-code your CRM' debate
---
Travis: I'm a HubSpot partner, and I spend all my time in HubSpot.
Aaron VanStone: Yep
Travis: and then I [00:21:00] hear people say, "Oh, we're just gonna vibe code our CRM." I'm just like, that's, that's great, but now how are you gonna train people? How are you gonna hire
Aaron VanStone: Yep
Travis: and say, "Hey, do you have experience with this thing we built and nobody else has ever used before?"
Aaron VanStone: Mm-hmm.
Travis: And then how are you gonna hi- you know, are you gonna now hire a technical team to make sure it's up all the time?
And I mean, I think the vibe coding is cool and, you know, being able to build something like that, but I think the long-term implications and from a business perspective, it's kinda shortsighted you have to hire somebody and... 'Cause I was a recruiter too, and I remember, you know, recruiting for Microsoft and other, Comcast and other companies, and they had a
Aaron VanStone: Right
Travis: set of, of technology they were using. And it was common to have people with that experience, right? And then I would talk to somebody in, in a phone interview, and they had been working for five years at this company that developed their own thing, and I'm like, "Are you gonna be lost coming here and working on this after you've just been in that [00:22:00] thing that nobody
Aaron VanStone: Right
Travis: on?" and so that kind of disqualified people. ~Um, ~and so I think that could also happen as people work at these companies where they've vibe coded something, they try to go work somewhere else. They don't have the experience of working in the
Aaron VanStone: Yeah.
Travis: else is working in. So
Aaron VanStone: No, that, that's, that's a really good point, and that's something that I never-- that never crossed my mind. So, so when I was building our CRM, ~um,~ and I mean, and it's, it's simple right now. A- and I don't think it's going to be incredibly complex because it's, it's kind of one straight channel at this point.
Now obviously, if things grow, then, then things get a little bit different. But the, the pro in my mind is, is that you can control your own data, you can control... Not like you can't in, you know, e- every situation's different, but we can control-- We own our own data, we control it, and then we can, ~um,~ y- we can quickly, because these tools are so good, that if I need a [00:23:00] report that's never been written before, well, after about 15 minutes of, you know, vibe coding, I have that button to run that report, and now I have that flexibility of doing that where I don't...
So for me, I don't know HubSpot, and so I've never, I never worked on that. And I know Salesforce like a little bit, but I didn't wanna be beholden to Salesforce. So it kind of, it kinda made sense to go your own direction where you're building exactly how you want it. But you make a good point because that's how I want it.
Travis: Mm-hmm
Aaron VanStone: That may like, what- whatever works in this crazy thing may not be
Travis: Yeah
Aaron VanStone: what, what, what works for, like if you came in and looked at like, "W- what are you doing?" ~Uh, ~you know, "Where is, where are all these standard things that clearly you don't even, are looking at?" So that could be... I could see that being a, a pitfall down the road.
I'll... May- maybe we'll have to do a, ~uh,~ 2.0 [00:24:00] conversation to say a year later,
Travis: Yeah. Yeah.
Aaron VanStone: worked and what didn't.
Travis: Exactly. Yeah.
Aaron VanStone: Yeah.
Travis: And, and I think maybe moving forward, people may have to, like, vibe code their things, but it off of a more common UI, right? Of
Aaron VanStone: Right
Travis: "Hey, we're not gonna copy HubSpot, but let's create the functionality and kind of the UI that nav and, and kind of the, way it's organized,
Aaron VanStone: Yep
Travis: ours in a similar way so that when people get in, if they've used HubSpot, that could be a hiring, know, piece of, 'Hey, have you used HubSpot?'
'Yeah.' 'Okay. Ours pretty much is laid out the same way. We don't have the bells and whistles, and HubSpot's changing all the time, so it's not one-to-one. But
Aaron VanStone: Right
Travis: to navigate HubSpot and how everything, the objects are all organized and associated, then you could use ours.'" And I think that's what people may have to start doing, is just finding a, a, a common of- often used, ~um,~ piece, ~um,~ and then [00:25:00] just try to be as similar to that as possible for
Aaron VanStone: Yeah, it, it's, it's a fascinating... It, it is fascinating to think about that we're so quickly heading into an area where everybody is just so individualized.
Travis: Mm.
Aaron VanStone: like there's some pros to it, but there's also some cons, and there's a reason why, like, ~uh,~ in some instances, like are we trying to reinvent the wheel where some of these things work so well because you could depend on them and, you know, they were truly, ~uh,~ efficient on how they were laid out and h- and how the, the data flows.
So yeah, it, it... That's gonna be a kind of a weird offshoot from all this
Travis: Yeah. Yeah, 'cause now you're building stuff, and not only now do you have to learn Claude Code and build the thing you're building and become an expert in that, but then if you
Aaron VanStone: Right
Travis: a CRM, you kind of have to become a CRM expert, and then an expert on
Aaron VanStone: Right.
Travis: an expert on this
Aaron VanStone: Yep
Travis: And 'cause I've found with AI, it's great, but if you really [00:26:00] don't know, like, the ins and outs of that area that you're trying to, to explore, right, that topic or whatever, gonna pick up where it's wrong. ~Um, ~and when I do the things that I know well and I pick up what's wrong, it kinda be-- I'm kinda like, "Oh, man," like, it's wrong a lot. how much of this other stuff that I don't know it very well is wrong and I'm going with it, right? And so I think that's, ~uh... ~as, as great as it is, I think there's still, like, little nuances that if you don't know that area, you won't pick up on those things that could come back and derail things later.
Aaron VanStone: You know, that was a, and, and I don't know if this is a, this may g- be a little bit more apples to oranges, but it, it made me think of, so
The observability fix: 4 hours to 90 seconds
---
Aaron VanStone: when building ScriptSafe, one of the most important things that Google told us that they want to know, they, they wanna make sure that we are monitoring web, pharmacy websites afterwards so they don't start, they don't get approved for [00:27:00] accreditation and then start selling something illegal afterwards, right?
So, so then we start building this monitoring program, and for one of the, for one of the pharmacies that are, that's a client already, it was taking well over two hours. Actually, it started at four hours to scan the website for a night. Like, well, that's not sustainable. You can't, you can't have thousands of clients, hopefully, and, and each take that long.
So w- it was a concerted effort to push to get it, but every time, every night, there was a new set of bugs happening. And, and then, but it seemed like after, after I would try to fix it, it seemed like it was guessing what the bugs were or, or what the issue was. So come to find out, and this is stuff like, again, because I don't know kind of how all of these things are built, that there was no [00:28:00] observe layer, ~um,~ in this program.
So it had no way to actually look at what was happening and see what was truly causing the errors. So it was just running, and then it was just making guesses of like, "Well, this popped up. I don't know why." So then spent a ton of time on the observability layer of it just so that it could truly analyze.
And then once that happened, then we were off and running because then it knew exactly what the issues were and we, the system we built wasn't good enough, so we had to redo the entire design of it. And then now we're getting scans for even the large pharmacies, the most is like 25 minutes. A lot of them, they can be done in, in under 90 seconds.
Travis: Wow
Aaron VanStone: but it's, but it's, it's finding those if, if you don't know where to look and you're assuming that, like it's smart enough to know that. Well, I never, I, I never told it to build in that piece [00:29:00] of it, so it didn't. And, and then you're just, you're kind of, it seemed like, ~uh,~ for a good month there, I was just trying to chase my own tail, ~uh,~ until we figured out a better way to do it.
Travis: Yeah. And if you're assuming and then it's assuming, like that,
Aaron VanStone: That's right. That's, that's an issue. That's an issue.
Travis: Mm-hmm. Yeah. And, and d-
Starting from zero coding experience
---
Travis: before you started building all of this in CloudCode, did you have any coding experience, or have you done much in it in the past?
Aaron VanStone: y- you know, I took some business classes at, ~uh,~ you know, WSU, God, you know, years ago. ~Uh, ~I mean, we're-- I'm still part of that generation that growing up we didn't have internet. You know, cell phones came out when I was in college, so,
Travis: Yep
Aaron VanStone: the most part, you know, like being widely available and whatnot.
So, ~um, uh, ~you know, I understand. I've, I've built things, and I enjoy building things, so I, I understand the concept of, you know, what it takes to, to build something, even [00:30:00] if it's with your hands. So that has helped a little bit. But I mean, really for the most part, like some of the concepts I, I know, but r- for the most part, like I, I had zero experience.
I mean, it was just, it was just kind of seeing what would work and, and then being that, like I said, now that I've been doing it over, you know, it feels like every day for the last four or five months, you know, what I'm doing now, I'm just so much more efficient because, you know, it's, it's... Y- you just learn
Travis: Yeah,
Aaron VanStone: A- a- after a while, ~uh,~ "Oh, that didn't work, so I'm not gonna do that again 'cause that leads me down that, that r- that, ~uh,~ that rabbit hole that I don't wanna go back into, so I'm gonna tell it right away not to do that." And, you know, you just, you know, things you learn along the way
Travis: Mm-hmm. Yeah, exactly. And I think it's just, it's super important to, ~uh,~ be able to have the time, the deep, deep work time, the deep concentration
Aaron VanStone: Yep
Travis: to start to go down that and [00:31:00] just have the time to spend to, ~uh,~ to test and learn and, and figure it out, right? And I think it's hard for a lot of people our time's so fragmented to really... think that's why a lot of people feel like they're behind with AI because they haven't, they-- not many people have that time to,
Aaron VanStone: No
Travis: dig in and learn like you need to learn. ~Um,~
Aaron VanStone: You know, and, and what's funny to me is, is that I,
'If you don't know where to start, ask it' + the hotel scanner
---
Aaron VanStone: I went to a, a conference, ~um,~ about-- I've been to a couple conferences on AI and whatnot, and, and just to hear what other people are doing. And, and one of the-- And there were a lot of people in the room, the one, one conference that I went to, a lot of people in the room that were just like, they didn't even know where to start.
And so someone on the stage just said: "Look, if you don't know where to start, ask it.
Travis: Yeah,
Aaron VanStone: Where should I start?"
Travis: Yeah
Aaron VanStone: And, and then that's, and that's where it goes. And, ~uh,~ what's funny is, is the very first thing that I built using Claude was there's this hotel locally, ~um,~ I live in Spokane, and so two hours away [00:32:00] called The Post Hotel out of Leavenworth, is a really nice hotel, but it's really tough to get into.
So you have to book almost four or five months in advance, and to get times availability is tough. So I asked it: "Can you build a scanner that will actually scan availability and run twice a day and give me emails that, you know, oh, this opening popped up, you, you know, here, here you go?" And within about 15 minutes it was working.
Travis: Nice.
Aaron VanStone: And so then you kinda... And, and then, you know, it didn't work perfect, so like: "Ah, can you adjust this?" And then it updates, and then it got better. And, and all of a sudden like, wow, you, you really-- So instead of me having to go on the website and look for those dates every day constantly, now I have a program doing it, and I don't have to worry about that anymore.
So it's kind of like a simple concept of there's been times going through this [00:33:00] process where I'm doing something manually, and you have to kinda break that habit of like, no,
Travis: Yeah.
Aaron VanStone: you're d- There's probably a easier way to do this, to use the tools that you have. So to like, you know, your case in point, ~uh,~ s- have these tools do all the background manual work so that you can spend more time doing the actual podcast, or you can spend more time on your actual nonprofit, you know, making it grow instead of just trying to, the minutiae of just getting the, you know, the, the, all the, ~uh,~ the labor-intensive manual task, ~uh,~ taking up all your time.
Travis: E- exactly. Yep. Yeah, and when you're a one-person show, you know, you don't really have that time to, to do those things that you need to do
Aaron VanStone: No.
Travis: the business, right? And,
Aaron VanStone: Right
Travis: So yeah, and I think, you know, it also me, ~uh,~ think about when, when we did meet at Starbucks and were talking about this, ~um,~ I think I [00:34:00] had mentioned that, you know,
Why now is the time to learn — even if you've been laid off
---
Travis: being laid off is awful and it
Aaron VanStone: Mm-hmm.
Travis: and it's never fun, but... And there's a lot of people that have been laid off lately. ~Um, ~but I think there's really never been a better time to be laid off. Ideally, the-- Especially if you get a severance package,
Aaron VanStone: Yeah. A
Travis: paid to learn
Aaron VanStone: little bit of a runway
Travis: So now you have that time that, that the
Aaron VanStone: Yeah
Travis: are working where you left probably don't have the time to really learn this stuff.
And so actually you're getting that time that's super important right now, super valuable, could, could really propel the next five to 10 years of your career. And when... As you get back into to getting back into the market and getting back into the job market, gonna be light years ahead of the people at the company you just left.
And so I, I think as much as it sucks, I think there is a silver lining right now for people who have been laid off to really dig into this stuff because they have the time that they didn't have before. So
Aaron VanStone: [00:35:00] I, I couldn't agree more and 'cause it's, it's, you know, I've, I've been there before, ~uh,~ and it's, it, it, it's, you know, no, it, I mean, obviously it's no fun and then, and then it's just this, you know, where do I go next type feeling. But, you know, being on this side of things, and to our point earlier, that $100 or $200 that you spend for a subscription to, for, like what I use, I use Claude primarily.
So a $1 to $200 subscription, the value you're getting out of that is absolutely incredible. And so, you know, if, if it's, if, if I were to give, you know, myself advice, if I were into that, if it, if I was in that situation and I didn't know these tools, I would get on and for the first week I would try to figure out how to build a website and what are all the connectors and all the different things that I do to build my own website and, and how all of that works.
That's like, that's the [00:36:00] first thing, and then you can see live results from that. And then maybe the next time, maybe I try to figure out how to build a database and collect a, you know, a program that collects data and that kind of stuff. Just these different skills that companies need and understand how these AI models, ~uh,~ can help you build those because, you know, it's, it's an, it...
There is-- They're, they're incredibly powerful, but my sense is, is that even though there's a small percentage of us that are talking about this, the far majority of the people out there, they, they ha- they, they'll use ChatGPT to give them a recipe for tonight's dinner, but they're not using it for, for any of this and they don't know how to set up, they don't know how to set up, ~uh,~ their business for it because they don't know where to start.
So if you can figure out how these [00:37:00] tools work so that you can walk into a business and say, "Ooh, I can make this so much more efficient. We can put a brand new website up and we can have this great sales funnel flow for you," and then have that integrated into your entire like say attire shop or whatever, ~um,~ my goodness, you just created yourself a job and you just probab- and you probably just, ~uh,~ created yourself a job that pays more because they need you.
Travis: Yep. Exactly.
Aaron VanStone: they can't do this without you.
Travis: Exactly. Yeah. And, and I think, you know, when you get laid off, your confidence is down, you start to do some things, you feel imposter syndrome. You don't really wanna go out on social when you need to be on
Aaron VanStone: Yep.
Travis: engaging with people 'cause you
Aaron VanStone: Yep.
Travis: you know, your confidence is rock bottom.
But start to build a few of these things, and you actually see what's happening, your confidence starts to grow. And then, you know, I, I still feel like I'm super behind and, and I know I am on some things, right, with AI. ~Um, ~[00:38:00] but then I go out and I talk to people, and I start talking about this stuff, and it's like I'm talking in another language, and I'm like,
Aaron VanStone: Yes
Travis: I, maybe I am further ahead than I think I
Aaron VanStone: Right.
Travis: know, more than the general population." And so, ~um,~ so yeah, I kinda f- get the confidence back of like, I'm behind, but I'm in a better spot than a lot of people right now.
Aaron VanStone: I, I mean, you know, like, you know, this, th- here's the plug for your podcast. If you're listening to this podcast, you are way farther ahead
Travis: Mm-hmm.
Aaron VanStone: than anybody else out there.
Travis: Yeah, absolutely. Absolutely. ~Um, ~cool. I know we're, we're coming up on time. ~Um, ~and I, I did wanna ask you quickly, ~um,~ and hopefully we have enough time to, to squeeze it in, but you, you built a website. I hear a lot about people building, building websites. I, I'm getting ready... I wanna redo my, my nonprofit's website 'cause I had to build it quickly, ~um,~ and spin it up. ~Um, ~when, when you or other people are building these things in Claude Code, are you building it essentially in something like WordPress [00:39:00] hosting it somewhere like Squarespace or Google or any kind of the hosting programs? Or how, how does that even h- work with Claude Code? 'Cause you gotta have it go somewhere, right?
Aaron VanStone: Right. Right. So, so here's the,
The starter stack: Vercel, GitHub, Supabase
---
Aaron VanStone: here's the flow that, that I have, that I have done for mine. So Claude has what it calls connectors. So you can... There's different, ~uh,~ other companies out there that it will connect directly to that you can use so that when you are doing Claude Code, ~uh,~ it will automatically update your website and different things like that.
So the three main things that, three, ~uh,~ that I use is Vercel, which, ~uh,~ is actually the, ~um,~ company that puts the website on the web, ~uh,~ and it will, ~uh,~ it, it will update that. GitHub is where your code stays so that your s- your code is on the web so that it can, ~um,~ [00:40:00] drive to Vercel. ~Um, ~so you need those two kind of working together.
And then the final thing is Supabase, which is your database. So those three are... You can, you can have subscriptions, all that kind of stuff for probably, I think it's... My guess is I, I think you can get all of that monthly for under $30. And I think one of-- I, I think they may even have some free options as well.
Once you have GitHub, Vercel, and Supabase, you can do anything. So you can quickly go into Claude Code, build something, ~um,~ and when you link your keys together, it takes, you know, probably about 10 or 15 minutes to get everything linked, and then you're off and running. ~Um, ~so that, that's the, that's the flow that you use.
~Um, ~
Design: Manus, Claude Design & brand systems
---
Aaron VanStone: I was using a, and you know this, ~um,~ a AI kind of like tool called Manus. ~Um, ~and Manus was really good at [00:41:00] design work because for the longest time, Claude, you could, you could see a Claude-designed website or flyer a mile away on the boxes that it used and things like that. And so we did-- I didn't want that.
I wanted something a little bit more, ~um,~ you, you know, unique. So Manus, ~uh,~ was, was very good for that. Problem is, is that Manus is fantastic on design and it is awful on actual coding. So it, it just... It does not. It's, it's errors galore. But, ~um,~ Claude released Claude Design. So use Claude Design and it's...
You have to do it on the web. So how I did it is that you would go to Claude Design and you would have it build the entire like UI of the front-facing portion of the website so that it looked really good, and then, and then you took those files to Claude Code and say, "All right, now build it ~Um, ~and then one last [00:42:00] tip, 'cause I think this is probably one of the better tips.
Two big tips to finish
---
Aaron VanStone: The two big tips that I would say for this area, one, like I've said before, install Codex so that Codex is always reviewing what your code is. Every time, even with Fable, Codex will find things that needs to be adjusted logic-wise from Fable. So when those two agree, you have something pretty solid. And two, build...
When you're doing a website, build a brand system. So build it, so like for your nonprofit, say, "I like these three colors. This is my logo. This is how I want-- This is the font I want. This is the type of look that I wanna have." Spend some time doing that, because then when you're building the website, even in Claude Design or whatever, say it has to follow the rules of the brand system, and then everything [00:43:00] looks unique all the way through, or, or looks, looks consistent all the way through.
Travis: Yeah
Aaron VanStone: and then w-when you do that, and especially like with Fable, you may not even need design anymore because Fable is so good at it, and you just tell it, "Follow my brand system. This is, this is what I need in the website," and done.
Travis: Nice. And, and when you're doing
Aaron VanStone: Pretty cool
Travis: ~um, ~you know, I think with traditional websites in the past, you'd go and y- when you're working with a design team, you would go to other websites and you'd pull out things you like from those. Can you still
Aaron VanStone: Yep. Yep
Travis: and just say, "Hey, I like the way this navigation is on this site. I like the way this thing's structured on this site," and kind of pull from different sites and, and put that in? And will it be able to then take that and, and help you build that brand?
Aaron VanStone: So the website that I built yesterday, and this is for a separate project, and this is for another company I'm working with, it was just a unique one that we're just running, ~uh,~ kind of like a there's gonna be a [00:44:00] flyer, and then you're gonna take a scan a QR code to see this website. Just kind of like an informational piece, side piece.
~Um, ~one of the things I told it was is that this needs to be consistent with the actual company's website. So it matched the branding of that and applied it all the way through to this new website. So if you tell it, "Match the branding of I really like, I really like this Nike site right here, can you make it look like that?"
Specifically give it that and tell the things that you really like. "I really liked how this font popped off the page, and I really liked this color scheme." Tell it that, and you'll be amazed on how all of a sudden, just those details alone will give you that, that type of look right there.
Travis: Nice. 'Cause yeah, I can go out to the site and look at the code behind it and
Aaron VanStone: Right.
Travis: how it's
Aaron VanStone: Yeah. Yep, yep, yep. Yeah. Yeah, that, that's exactly it, yeah
Travis: Nice. Well, yeah, really appreciate you [00:45:00] coming on. It was a great conversation, and, ~uh,~
Aaron VanStone: Yeah, it's been fun.
Travis: two at some point
Aaron VanStone: Yeah
Travis: to get further down and I am further down and, and kinda graduating from building these little apps and
Aaron VanStone: Right.
Travis: ~Um, ~and kinda compare notes and, and see how, especially with Fable, ~uh,~ and whatever might be out in three months, ~um,~
Aaron VanStone: Right.
Travis: how that's affected things, right?
Aaron VanStone: Yeah, let's, let's, let's get back, you know... I mean, what's funny is, is that I was saying, "Oh, maybe in a year," but the way things change, we could probably get back in two weeks and the entire landscape has changed and we have new stuff to talk about.
Travis: Exact- because yeah, I think we just met last month, and now Fable's out and things are already different, right?
Aaron VanStone: That's exactly right. That's exactly right. So, so yeah. Well, thank you so much for having me on. It's, it was a great conversation.
Travis: Yeah. Thanks for coming.
And that's it for this episode. Before you go, do me a favor. If you got value out of this episode, leave a quick review. It's the best way to help the show get in front of more founders and revenue leaders who wanna make AI actually [00:46:00] work for their business. See you next time