Pod Insights How I AI How a 91-year-old vibe coded a complex event management system using Claude and Replit | John Blackman About this episode John Blackman, a 91-year-old retired electrical engineer, shares how he used Claude and Replit to build a complex application for his church’s community service events—with no prior software development experience and for less than $350. His app allows event organizers to create events, recruit volunteers, and manage sign-ups, with a standout feature for organizing free oil changes for participants.
What you’ll learn:
How John used Claude to create detailed product requirements and user storiesJohn’s philosophy on embracing new technology throughout his careerThe exact process for integrating third-party APIs (like VIN lookup for oil changes) with minimal technical knowledgeHow he automated report generation for volunteer management and resource planningHow the software generates personalized Impact Passports for event participantsWhy letting AI build without preconceived notions of “correct” implementation can lead to faster, more functional resultsHow to troubleshoot common development-to-production issues when working with AI coding tools—
Brought to you by:
WorkOS https://workos.com/?utm_source=lennys_howiai&utm_medium=podcast&utm_campaign=q22025—Make your app enterprise-ready today
Orkes https://www.orkes.io/—The enterprise platform for reliable applications and agentic workflows
—
Where to find John Blackman:
Website: http://johnbeng.com/
—
Where to find Claire Vo:
ChatPRD: https://www.chatprd.ai/
Website: https://clairevo.com/
LinkedIn: https://www.linkedin.com/in/clairevo/
X: https://x.com/clairevo
—
In this episode, we cover:
(00:00) Introduction to John Blackman and his background
(02:55) John’s impressive career
(03:59) How the church project started
(05:06) Using Claude to create a development roadmap and requirements document
(07:29) The concept of the Impact Passport for event participants
(08:57) Generating user stories and requirements with Claude
(10:32) The multi-tenant architecture with system and local church administrators
(12:54) Building the application with Replit
(13:32) Demo of the administrator interface and event management features
(17:56) Specialized reports for different services (food pantry, vision center, oil changes)
(20:30) The participant registration flow with QR code scanning
(21:55) Adding new features like volunteer name tag generation
(24:40) Troubleshooting AI “rabbit trails” during development
(26:09) Challenges moving from development to production
(27:13) John’s lack of coding experience
(29:42) The advantage of having no preconceived notions about implementation
(30:25) Total development costs and timeline
(31:31) Impact and reception from the church community
(32:42) Lightning round and final thoughts
—
Tools referenced:
• Claude: https://claude.ai/
• Replit: https://replit.com/
• SendGrid: https://sendgrid.com/
• AutoCAD: https://www.autodesk.com/products/autocad/
—
Other references:
• OpenAI API: https://openai.com/api/
• VIN (vehicle identification number): https://en.wikipedia.org/wiki/Vehicle_identification_number
• Multi-tenant architecture: https://en.wikipedia.org/wiki/Multitenancy
• Role-based access control: https://en.wikipedia.org/wiki/Role-based_access_control
• Excel: https://www.microsoft.com/en-us/microsoft-365/excel
• Docusign: https://www.docusign.com/
—
Production and marketing by https://penname.co/. For inquiries about sponsoring the podcast, email jordan@penname.co.
Transcript Speaker names marked ? are inferred from speaking time and the show’s known hosts — they are not voice-verified.
0:00 Speaker 00 We do Impact Weekends at our church. We go to a local church and provide free haircuts, free eyeglasses, free car wash, free food and everything. I handle registrations for these events. And so I said, it'd be nice to have that computer somehow. So I wrote up a kind of an outline of what I wanted to do. We sent it to Claude, told Claude we wanted for Replit. Then we had him write our program and we sent it to Replit then to do the program.
0:25 Claire Vo? If you all told me the story correctly, John, you and your other grandson did this into the wee hours of the night.
0:32 Speaker 00 We start at 10 and finish about 3 o'clock in the morning.
0:35 Claire Vo? It's beautiful. You've got beautiful navigation. It's easy to read. It's simple to navigate.
0:41 Speaker 00 So I have control of all of the churches. And here I see all of the participants that have registered. I have the services that are available. I have reports and I can print this report out beforehand to know what people are coming.
0:54 Claire Vo? Do you have any wisdom or advice for us as folks in our professional careers are facing this technology change?
1:00 Speaker 00 It's just like AutoCAD. A lot of my friends didn't want to learn AutoCAD. And so when I retired in 94, I still was working in 2018. I was still having fun.
1:09 Claire Vo? So that's another reason to learn this technology, because if you learn it, you can be having fun well into your 70s, 80s and 90s.
1:21 Claire Vo? Welcome back to How I AI. I'm Clarovo, product leader and AI obsessive here on a mission to help you build better with these new tools. I'm just going to get to the punchline. Today, we have John Blackman, a 91-year-old vibe coding grandpa. He used Claude and Replit to build a very complicated, very impressive app for his church, and he's going to show us how he did it. Let's get to it. This episode is brought to you by WorkOS. AI has already changed how we work. Tools are helping teams write better code, analyze customer data and even handle support tickets automatically. But there's a catch. These tools only work well when they have deep access to company systems. Your copilot needs to see your entire code base.
2:06 Claire Vo? Your chatbot needs to search across internal docs. And for enterprise buyers, that raises serious security concerns. That's why these apps face intense IT scrutiny from day one. To pass, they need secure authentication, access controls, audit logs, the whole suite of enterprise features. Building all that from scratch? It's a massive lift. That's where WorkOS comes in. WorkOS gives you drop-in APIs for enterprise features so your app can become enterprise-ready and scale up market faster. Think of it like Stripe for enterprise features. OpenAI, Perplexity, and Cursor are already using WorkOS to move faster and meet enterprise demands.
2:48 Claire Vo? Join them and hundreds of other industry leaders at workos.com. Start building today. So I don't usually start this podcast with bios, but John, yours is too good to not give a little time to. So let me embarrass you a little bit with a quick bio of your experience. So you started as an electrical engineer at Kansas City Power and Light, and then you left to run a hardware store in Oklahoma. While you were in Oklahoma, you earned your airplane mechanic certification. Then you returned to Kansas City and rejoined engineering, becoming the first in your department to learn AutoCAD in the 1980s, and you were eventually training others.
3:31 Claire Vo? You worked on one of the first underground fiber optic projects in the US. You were brought out of retirement to help launch Google Fiber in Provo and Kansas. And then at 91, you are vibe coding software using AI agents. Is there anything you haven't tried?
3:48 Speaker 02 Not yet.
3:49 Claire Vo? Not yet. Oh, and I forgot the cherry on top. You've owned Bitcoin since 2018.
3:55 Speaker 02 You've got some Bitcoin, right? That's right. Yep.
3:59 Claire Vo? So you have the epitome of a growth mindset. You've had such an amazing career, learned so many different fields and practices, and now you are into AI. How did this all start?
4:15 Speaker 00 Well, it's Brett's fault. My other grandson. Well, I was talking about, we do Impact Weekends at our church, which we go to a local church and provide free haircuts, free eyeglasses, free car wash, free food and everything. It's kind of a ministry-type thing. I handle registrations for these events. And so I said, it'd be nice to have that in the computer somehow. And so he said, well, let's do it. So I wrote up a kind of an outline of what I wanted to do. We sent it to Claude and then Claude turned around and we said we were going to send it to Replit,
4:57 Speaker 00 told Claude we were gonna, we wanted for Replit. Then we had Rip him write a program and we sent it to Replit then to do the program.
5:06 Claire Vo? Okay, so you had this idea to support your church and your ministry services and you just needed something that was more scalable than paper. And you did what all my peers in technology are now doing, which is ask Claude to create what we in the business call a requirements document and then you sent it over to Replit. So Brandon, can you help us dive into the actual Claude chat and where this all got started? And I, I think if you all told me the story correctly, John, you and your other grandson did this into the wee hours of the, of the night. So you got started with Claude and then did it till, you know, one in the morning or maybe even longer than that.
5:48 Speaker 00 We started 10 and finished about three o'clock in the morning. And then we started the next day at 10 until about five o'clock that evening.
5:56 Claire Vo? Well, you are in good company because I think 10 to three in the morning is every Vive coder's schedules. You're good. Okay, so tell me how, you know, you started here. It sounds like you started by asking if you could plan a development roadmap. So you went right, right to it.
6:14 Speaker 02 And Grandpa, what's cool about this I've picked up on is that you have things like in here, like if you need more information from me, ask me the questions right away. Like you were telling it exactly what it should be doing. And then it started just going from there and you started providing information.
6:35 Claire Vo? Yeah, so you started at the top with a very general query, which was, how do I create a roadmap, basically? And then you got the chat to actually ask you enough questions about your specific project that it could do a good job. So you have a draft of what you want the application to do. And in specific, you say it wants to be built on Replit. How did you find out about Replit? Why did you pick that application?
7:00 Speaker 00 That was suggestion by Brett.
7:03 Speaker 02 Yep.
7:03 Claire Vo? All right. My, my Replit friends, do not underestimate grandson influencer marketing for your agents. Okay. So he said, let's just build it with Replit. And so you instructed the system that. And then, you know, did you list out all these, these ideas here? Were these from you two?
7:22 Speaker 00 Yes, we, I typed them up in a word document and then we sent that to Claude.
7:28 Claire Vo? Great. So this is a very structured, step by step, what we would call kind of user journey or flow. So you're talking about registration, what data you need, and then what kind of services they can do. And these are all wonderful services that you can get in your community. You know, haircut, dental service, those kinds of things. And then how did you come up with this idea of an impact passport?
7:54 Speaker 00 I saw that at another church one time. They had, they did a lot differently than I do it. But anyway, they had a little, what, a piece of paper that the people would carry around with them to this event. And so we do that too, but we do it by, we would write it out by hand. And so the, I wanted something that they, we could print out and then they could bring with them to the event.
8:19 Speaker 02 Yep. And so they'll sign up for something online. They'd be like, okay, you signed up for the haircut and you signed up for the face painting. And then you signed up for the health clinic and then they'll get their passport and then they know what stations to kind of go to. And like you mentioned, that was all done by hand beforehand. So.
8:36 Speaker 00 Right.
8:36 Claire Vo? So what you're trying to do here is you have a bunch of manual handwritten processes that you're doing to run this event. And you're asking Claude, how do I turn this in to software? And how, how could I do this both on the volunteer side, on the ministry side, as well as the folks coming to receive services? And so I see in this chat, you're getting a, a nice back and forth between you and Claude. And then let's see what the roadmap looked like.
9:08 Claire Vo? All right, so you got the core features for an MVP, the first piece of it. The admin interface, passport generation, as we spoke about, and then data management, all that good stuff, and a bunch of development phases. And then it looks like on the side, Brandon, it also generated user stories and requirements.
9:31 Speaker 02 It did, yes. And I see here it said QR code. Did you know it was going to do a QR code?
9:37 Speaker 00 Well, I'd seen this done at this other church with, for their passport. So I wanted to do that in ours because it would, it'd make it a lot easier because everybody has an iPhone now. And so they can just scan that code and then it brings up the registration form for them to fill out.
9:54 Claire Vo? So you were taking inspiration, not just from your manual processes, but from ways you've seen other churches do this work. You sort of brought that all together, used chat to put that in an organized structure. And then we're looking at these user stories. Brandon, do you want to walk through kind of some of the key user stories here?
10:13 Speaker 02 Yeah, so he's got the registration chunk of it right here, which a lot of this would be driven from someone who's typically responding from like a Facebook ad or something like that, correct?
10:26 Speaker 00 So Facebook, we have brochures that we hand out, about 5,000 in the neighborhood.
10:31 Speaker 02 Yep. And this is the part that honestly, even to this day, I still can't wrap my head around in terms of just the complexity of the multi-tenant setup that he has here. So there's an admin interface and that has to have lots of different levels because, and you keep me honest here, Grandpa, you've got the overall impact organizers, which sit at one admin level.
10:56 Speaker 00 Which is like we call system admins.
10:58 Speaker 02 System admins. And then beneath that, you've then got all of these churches.
11:03 Speaker 00 Individual or local admins.
11:05 Speaker 02 That then have their own admins and their own logins and their own data. And visibility can go up and down. So the, the impact admins, the system admins can see all the way down, but the individual church admins can only see their own church and their own events.
11:21 Speaker 00 And the system can actually approves all of the admin administrators.
11:27 Claire Vo? For those that are not on the YouTube or the video, I am smiling ear to ear because you're building this complex, what we call roles-based access system, multi-tenant, Brandon, as you said.
11:40 Speaker 00 Yep.
11:40 Claire Vo? Complex piece of software with admin backend functions, you know, front-end, you know, consumer or participant-facing functions, all from a paper process and something that you just want to make better for your community. And so I'm seeing those two kind of like major users. And then as we go down, we're seeing these user stories, which again are, I love user stories as a product person because it really lets you describe how you want the software to be experienced from the user's point of view. And so, John, did you find these user stories were pretty accurate right out the, right out the gate?
12:22 Speaker 00 They were pretty accurate. I made a few changes to them as we went through the program, but mostly, yeah, that he had pretty well described what I was thinking.
12:33 Claire Vo? Yeah. And I even see here, you're not just looking at administrators and users, you're thinking, what would the pastor of the church want? What would the ministry leader want from all of this, all of this data? And you even have non-functional requirements. You are a very good product manager. This is better than some work I've seen in professional organizations. Okay. So you built this great set of product work with Claude. And then this is going to blow people's mind. You, you just took it to Replit Agents to start building it. So did you download, copy and paste this into Replit Agent? Do you remember what you said?
13:11 Speaker 00 I just took and copied what Claude had put together and put it into Replit, and then it started going blip, whoop, and there it was.
13:24 Claire Vo? Well, I have to tell you, that's the official AI noise.
13:29 Speaker 00 It was so fast, I couldn't believe it.
13:32 Claire Vo? Okay, so let's hop over to Replit and show a little bit of what you've built here. I'm going to pause really quickly. So I'm just looking at the side. You've got docs, migrations, you've got generated static assets here in these passports. This is a real meaty application here you built. So let's jump to the punchline. Let's do a little world tour of what you actually built. And then we can maybe go into some of the problems that you've been trying to solve lately with Replit.
14:07 Speaker 00 To start with, we go in, as a system administrator, I will go into this LBC multiple, which that's the system administrator. The LBC church is a local church. Testimony church is a local church and so forth. So anyway, I go into this and I sign in with an admin sign in. What this does, it brings me into the individual churches. So I have control of all the churches. And then when I go in here, then I can actually manage the event. So this way I could go into testimony church. And here I see all of the participants that have registered, which a person would do, the local church would do this, really. I have the services that are available.
14:51 Speaker 00 And these services can be turned on or off depending on whether they are provided or not. If they're not provided, you just click the button and it says no service for that event. And then I have reports. And the reports I have is demographics. That's everybody that's coming to church, coming to the event. It gives their name and it's name and address, phone number, address, and how many people are coming. Like if they're bringing kids and put, say, three kids or something, or three participants. And it's in an alphabetical order by last name. And I can print this report out beforehand to know what people are coming. Also, this demographics report is used by the pastor as a follow up
15:36 Speaker 00 for his ministry after the event's over. Then I have a service usage which tells me how many people use what service. So it will print out a list of how many use the pantry and how many haircuts and all this. So we kind of know how many, what services were better used than others. And then the oil change, it will show when the people register with their VIN number. So this gives me an oil report of what kind of oil, filters, and how many that I need for the project. So.
16:08 Speaker 02 And Grandpa, how did it, how do you know or how does it know what oil to buy?
16:13 Speaker 00 We use a VIN number search. Uh, when they fill out the registration, it will, it asks them for their VIN number. And then it searches for the VIN number and tells us what kind of filters you used, all this kind of thing.
16:27 Claire Vo? This episode is brought to you by Orkis, the company behind Open Source Conductor. The platform powering complex workflows and process orchestration for modern enterprise apps and agentic workflows. Legacy business process automation tools are breaking down. Siloed low-code platforms, outdated process management systems, and disconnected API management tools weren't built for today's event-driven, AI-powered, cloud-native world. Orkis changes that. With Orkis Conductor, you get a modern orchestration layer that scales with high reliability, supports both visual and code-first development, and brings human, AI, and systems together in real time.
17:09 Claire Vo? It's not just about tasks, it's about orchestrating everything. APIs, microservices, data pipelines, human-in-the-loop actions, and even autonomous agents. So build, test, and debug complex workflows with ease. Add human approvals, automate backend processes, and orchestrate agentic workflows at enterprise scale, all while maintaining enterprise-grade security, compliance, and observability. Whether you're modernizing legacy systems or scaling next-gen AI-driven apps, Orkis helps you go from idea to production fast. Orkis: orchestrate the future of work. Learn more and start building at orkes.io. That's O-R-K-E-S dot I-O.
17:57 Claire Vo? Okay, so I just have to pause for folks that are not on video to recap everything you vibe coded here. You have this admini- multi-administration levels. You have data models for different locations or churches. You have data models for different events. You have data models for different services that can be turned on or off. You have a list of participants and real-time reports of different ways those participants are either registered for events or coming to events. You have service-specific software, which will take, I'm just taking a pause here, a request for an oil change that comes
18:38 Claire Vo? with a VIN number, and it will go, I'm presuming, make some API call somewhere to tell you exactly what kind of oil that car uses. And then you generate a report for an event that you can export to Excel and use as a shopping list to go get oil or tell people what they need to bring.
18:57 Speaker 00 Right.
18:58 Claire Vo? Just that.
18:58 Speaker 00 Also, these reports, here's the oil changes report. So it says the person's name, and we have a check-in for the oil changes.
19:07 Claire Vo? Okay, and again, for folks that aren't looking, it's beautiful. You've got beautiful navigation. It's easy to read. It's simple to navigate. You have these little alerts, and we call them toasts in the app, that tell you when different actions happen. And then, I know there's a whole participant side of this, and maybe we can share some screenshots of that as well, where people can register and actually get a little printed-out passport that they can take to the, to the event.
19:36 Speaker 00 Yeah. Now, the other reports I have is for the food pantry. This orders all the food for the food pantry, and also for the lunch that we, we provide a free lunch. So it orders like hot dogs and hamburgers and buns and all that, that we need for the lunch. And then the vision center, when we go to the vision center, we need a report that will be used by the people running the vision center to know what the person's, that is coming to the vision center, what their age is. And everything, and that's filled out by this report. And then we have waivers that are signed by each individual so that the church is not responsible for anything.
20:16 Claire Vo? Yeah, so let's look at that piece from the participant perspective because you've built some things there that I think are pretty cool.
20:23 Speaker 00 Want to go through registration?
20:24 Claire Vo? Yeah, let's go through registration.
20:26 Speaker 00 We'll just use Test My Church because it's events. So you put in your name and then the phone number, you have to put your phone number in.
20:35 Claire Vo? Okay.
20:35 Speaker 00 And your address.
20:37 Claire Vo? So when you scan the QR code off a flyer or something like that, this is the form. Okay.
20:42 Speaker 00 Yeah.
20:43 Claire Vo? And does it, does it look good on a mobile phone?
20:45 Speaker 00 Yeah, it's pretty good. You can read it pretty easily.
20:50 Speaker 00 And then this one here is like the eye clinic.
20:53 Claire Vo? Yep.
20:53 Speaker 00 And then you say next, you can even do oil change here. So let's do the oil change so you can see how that works. It has a pro look up the oil change.
21:03 Speaker 02 And what, what API is it hitting, do you know?
21:06 Speaker 00 It's open, open AI API.
21:08 Speaker 02 Open AI, yeah, open AI's API.
21:10 Speaker 00 So here, see, it shows that it's a five, five W twenty oil, six quarts in the filter number. So then you say, okay. And then this is the waiver that they have to read through. And when it gets down to the bottom, it's all filled in automatically. And then you proceed to the signature.
21:26 Claire Vo? So watch out, watch out DocuSign. You have a signature capture flow here. Okay, you can do with your, your finger on your phone.
21:37 Speaker 00 And then you say, I accept the signature. And then you say, complete the registration.
21:42 Claire Vo? This is, this is very good. Okay, I could, I could look at this all day as somebody who built software, but let's go back to how you actually did this. So one of the pieces of the flow that I think is so interesting that people would love to get into is this idea of looking up the, this oil change. And so let's just say a feature like that, I'm sure wasn't the first thing that you thought of. It was probably something you added after. So how are you chatting with Replit to add these features? And how much of the work is it doing for you? How much are you researching outside? Can we look at some of those chats?
22:19 Speaker 00 Well, right now, what I'm working on is having a name tag for the staff at the event.
22:27 Speaker 00 That will be printed out on a Word document on labels. So that way, we could slip them into a plastic folder and had a lining around their neck. Because right now, what we do, we write it with a marker on a sticky one and stick it on their shirt. And sometimes people don't want that stuck on their shirt or blouse. And so this will hang around their neck in a folder. So this is what I'm working on right now.
22:54 Speaker 02 It looks like he uploaded this first with this prompt here. Claude then started to bring out some of these assets here. So there's the template that it used. And then eventually it started to code down here.
23:12 Speaker 00 Right.
23:13 Speaker 02 And then he brought it over to.
23:16 Speaker 00 Replit. Yeah.
23:18 Claire Vo? Okay, so you wanted to make these name tags. And so you had to add the idea of a volunteer into this already complex app. And so let's look through the chat that actually did that and how Replit built it. Okay, so it said, a couple of weeks ago, you said that you want to add a function that will populate a volunteer list. When a local administrator, so you're using your roles, fills an input page. And then it says all the information that you need in that. And so let's scroll down and show what Replit actually does for you. So with that paragraph of instructions there, you are adding a new table to your schema. So you're adding a data model. It's updating.
24:03 Claire Vo? So I mean, do you just sit here and watch this work? Are you totally fascinated by this?
24:08 Speaker 02 You just sit there watching it?
24:10 Speaker 00 I liked it. Yeah, I watch him. Sometimes he goes off on a rabbit trail and I have to back.
24:18 Claire Vo? And so this, you know, keeps going. What's interesting about this Replit agent implementation is you can see here, it's pretty independent. You're not doing a lot of back and forth with it. It's doing big chunks of work for you.
24:33 Speaker 00 He does.
24:34 Claire Vo? And then do you pull this up kind of locally in the browser and then check if it works and gives it? Give it feedback? How is that back and forth for you?
24:43 Speaker 00 When he gets to a point where he says, I think I got this, then I'll go back into the program and run it. Let's see if it works. Then if it doesn't work, I tell him.
24:54 Claire Vo? And so it made a mistake here where it says not all events. You want it for the local event. So it's going again and making these updates. And have you found that you've figured out how to prompt these agents? Are there any tricks you can share with us? Or do you just talk to it like you would anybody else?
25:11 Speaker 00 I just talk to it like it's a person, like you guys tease me about.
25:16 Speaker 02 Yeah, he, uh, he refers to, uh, the AI as a he. So the pronouns of he.
25:23 Claire Vo? Perfect. Chat PRD, my AI is a lady. So that's fine. And then you do do what all of us do with our AI, which is you just say, wait. And then I saw another one that said, stop.
25:36 Speaker 00 This is where he was going out on a rabbit trail. I said, wait a minute. Don't go that way. Stop.
25:42 Claire Vo? Yep. So one of the challenges, I think, working with these agents, which you're experiencing here is, as we see, it can do tons of work for you without intervention. But it's interesting that you can spot when you need to tell it to stop, or reset, or start over. It's pretty amazing to add these features. Now, I have to ask you, what was the most complicated thing, or the hardest thing that you built here?
26:09 Speaker 00 The hardest thing, and I finally got it about two days ago, was it always doesn't work in production, but works in development. And that was very frustrating. And so in my production, I was sending an email to the participant, informing them that they had registered for the event, and attached is their passport. So they'll print it out there at their home, rather than me having to do it at the event. In production, it would not attach that passport PDF file to the email. We worked on it for two or three days. And finally, the other day, we finally got it working. But the reason we did, we had
26:52 Speaker 00 to change the type of PDF file.
26:57 Speaker 00 Format that we were using. And so the passport that we send with the email looks different than a passport that we have in development.
27:07 Claire Vo? So I'm going to give you some, some real cred and credit here because if I just take a step back at what you've built, you know, everybody is going to listen to this and say, oh yeah, you can vibe code a registration app. It's just like a form. But if I'm looking at the complete, the complete nature of what you've built, you have a very complex application here that's serving many different users that has security needs, that has kind of opt-in and waivers. You're generating PDFs. You're generating Excel files. You're emailing those PDFs. You're generating reports. You're doing all these different kinds of software development,
27:50 Claire Vo? all wrapped in what I think is a beautiful UI. And just to confirm, have you ever coded software before?
27:58 Speaker 00 The only thing I did in AutoCAD where I worked at Paranite, we bought a program called Intellicad. It had logic files that we had to, they allowed us with open architecture to make our own logic files. So I did write logic files for that to show that conductors were attached to poles and so forth. And so I, but it wasn't really code like this. It was more like if then and plus, if this happens, do this, I'd say.
28:31 Claire Vo? But no TypeScript?
28:33 Speaker 00 No.
28:34 Claire Vo? So this is your first time writing TypeScript? And, you know, can you just talk us through some of the, you know, you said you use OpenAI APIs. Is Replit recommending what database to use? Is it recommending how to send emails? How much of that did you have to research yourself versus the agent telling you how it works?
28:55 Speaker 00 No, I don't know. I I've looked at it. I I've looked at the database to see, you know, what's what's in there, parts of it, but I don't know what it is.
29:06 Claire Vo? Got it. So when the agent tells you use this to store your data, or that to send your email, I think you're using SendGrid, you just take those recommendations and go. So for all of you coding agent builders out there, these out of the box integrations, John, I think make it simple for folks like you to add on new kind of technical capabilities without having to to research or make those decisions.
29:33 Speaker 00 Correct, yeah.
29:34 Speaker 02 That is, that's a good point. So I listen to a lot of other Vive coders, and that's probably a piece of advice I hear a lot, is that they will say, just be open minded. Like, when you go in, because I'm a I'm a technology guy, I work with software developers all day long, and we're very opinionated. Like, we know, like, how it should be done, and what tools it should be done with, and what the layers should look like.
29:58 Speaker 00 I don't know.
29:59 Speaker 02 And sometimes that can trip up the AI, because it's not going to naturally go that way, even if it's not the best way. And so it's interesting to kind of watch grandpa here just not have a single clue what it should look like, or what it should do. And AI just makes it work for him, even though no real software engineer might have done it that way.
30:18 Speaker 00 That's true.
30:19 Speaker 02 Yeah.
30:20 Claire Vo? I don't want to call you out too much, John, but I noticed you're on the free version of Claude. Have you paid any money to replicate for this app, or is it is on the free tier?
30:29 Speaker 00 I have spent, at this point, I think it's about $350.
30:33 Claire Vo? Okay, so $350.
30:36 Speaker 00 What amazed me, though, was those first two days, the program was basically running, and it cost me like $25 plus maybe 50 or something. Or, I think I I put a deal there, it cost me $171 when the program was running. And that's what Brett says, you know, it take his his programmers probably six months to do it. And we did it in two days.
31:02 Claire Vo? So I'm, I'm going to put some fear in the heart of the software engineers out there. You built some pretty impressive software, a couple of days, a couple of weeks, $170 in cost, free version of Claude, some, some grandson time, which I'm sure feels like you're getting paid, you're getting paid. What has the impact been on your church, on your community? What I can imagine here is you can just serve a lot more people, your volunteers are a lot more effective. Everybody's probably having a better time at these events because they're not stressed out about paperwork. What have you seen about the impact of what you built here?
31:45 Speaker 00 Well, to begin with, they couldn't believe it. They said, this blows me away. And so I've showed it to the mainly the pastors that are in charge of these events, but we have not implemented it yet because I keep having little things I keep adding to it. And some things are still not working exactly right. And so I'm still working on that, but they were really, they're ready to use it as soon as I can get it running right.
32:14 Claire Vo? Okay. Well, we're going to keep our fingers crossed that by the time this episode is released, you will have had your launch day for this impact platform.
32:24 Speaker 02 Well, that, that's great.
32:25 Claire Vo? Okay. And I bet you will recruit some, some volunteer developer assistance. And I even bet the Replit folks will help you out if we can get this in front of them. Okay. John, Brandon, this has been so much fun. Just to recap, you found a problem in your life and in your community that you thought could be made better with, your words, not mine, computers. You used Claude to create a requirements document, user stories, you downloaded those user stories, you put them in Replit Agent. As we've seen on screen, Replit Agent just through, through all the requirements. You have this complex piece
33:05 Claire Vo? of software. You're adding to it every day. You spent a couple hundred bucks. And I'm guessing you feel very empowered, like you've learned a lot in the past couple of months.
33:17 Speaker 00 Yes, I do. Yeah.
33:18 Claire Vo? Okay.
33:18 Speaker 00 It's been unbelievable.
33:20 Claire Vo? Great. Well, we are going to get you back. I know you're just going to code again till midnight. So we're going to get you back to your coding with a couple lightning round questions and we'll get you, get you out of here. So John, my first question to you is, we were joking before the show started, logging in to these applications is harder than actually building the applications themselves. But if you can make an ask to any of the kind of coding providers to Replit, to Claude, on things that they can make better for your experience, what would your ask be?
33:53 Speaker 00 The biggest problem I've had is the, it works great in development, but then when you have production where you actually want to use it, it doesn't transfer everything. For instance, right now, I have an OpenAI key. It keeps putting an old key in there, which is, I think he said it's stored in a cache somewhere. And it keeps putting this key instead of the right key. And I can't figure out why he keeps doing that. And so I haven't been able, that's why my VIN number does not work in production now. But it does work in development because the OpenAI key, it changes as we deploy it.
34:37 Claire Vo? All right. So you heard it here. Secrets management from development to production, still, still a problem. And Brandon, we had an example of something else that, that John wanted from, from Claude. Do you think we can pull that up or show that?
34:51 Speaker 02 Yeah, that's a great one. So we were, we were researching for the show, his, his chat history with Claude. And I don't know if you knew, did you know that the history was over here on the left hand?
35:04 Speaker 00 Well, I asked him where the history is, and he said he don't have it.
35:07 Speaker 02 Yeah. And so it'd be nice if Claude could say, well, here's all your chat history.
35:13 Speaker 00 He didn't tell.
35:16 Claire Vo? Yeah, so I think you had the sidebar because when we first started, you had the sidebar with all the history closed. And so the very obvious place to ask where my, where is my history is in the chat. But these chats, everybody, memory is key.
35:29 Speaker 00 It was over. I couldn't find it.
35:33 Claire Vo? Okay, so the preferred UX is just tell me in the chat and point me to, to my other chats. Okay, so we got good two pieces of feedback there. You know, my second question is what I love about you and are, you know, what you showed us here and your entire career is you have such a growth mindset around technology and embracing what's next. You know, when you, when CAD was coming out, you told me this story where people were really resistant and you leaned in. I'm curious, do you have any wisdom or advice for us as, you know, folks in our professional careers are facing this technology change that to some people can feel scary and to others feel really exciting. What's, what's your wisdom, having gone through this a couple of times?
36:17 Speaker 00 I was reaching this program to one of the people at one of these events. And she said, oh, I'm scared to death of AI. I said, well, why? She says, well, I just don't know what it's going to do. And I said, well, it does a lot of good things, but I'm sure it can do bad things. But I said, I think right now, if you figure out how to use it the correct way, it's going to help a lot of people. And I said, that's, that's why I found out here that this program, it, as a registration person, I don't have to do anything but stand there and hand out passports now. Where before I had to write all this stuff down by hand. And all the material that we have to order, it had to be done kind of by hand or by somebody else, you know?
37:01 Speaker 00 So now this, all of these reports will be there. And like the pastor follow-up ministry, now he'd have a complete report of everybody that attended with name and address and how many people attended. So he'd go call them and provide ministry for them. So I think it's been a good deal.
37:20 Claire Vo? I love that answer because I am like you, an optimist. I have to believe that if we put the ability of into the hands of people like you to build something, you will take it and build something amazing and good for your community and good for the people around you. And this is something that wouldn't have never existed. If these tools didn't exist, you would still be on paper. So we get to get a little bit of your imagination and your impact on the world because we have this new technology. That's really great, really great advice.
37:51 Speaker 00 It's just like AutoCAD. A lot of my friends didn't want to learn AutoCAD at the power line. And so when I retired in 94, I still was working in 2018.
38:04 Speaker 00 They had gone, I don't know what they did. I was still having fun.
38:09 Claire Vo? So, so that's another reason to learn, learn this technology because if you learn it, you can be having fun well into your 70s, 80s and 90s, building, building good stuff. There's a lot of longevity if you can learn more. Okay. And then we know that Replit is a he, so we know how you, how you speak to the AI. I am curious, when it doesn't listen, what do you do?
38:35 Speaker 00 I try, I try to keep going around the corner with it and say, try, in fact, I've told him several times, go back in history to a certain point and use this what used to work. And now you're saying it doesn't work. So go back and use that code. And sometimes it works because it was working before and then he broke it. I broke it somehow. He didn't do it. I told him, I told him to do something that made it break. And so, and crash. And so he would go back and pick that up and it would work again. And mainly that's what I've been doing. And then when worst comes to worst, I call Brandon and Brett.
39:15 Claire Vo? Okay, so when the AI does not work, go back to a checkpoint or call, phone a friend.
39:21 Speaker 00 Phone a friend, exactly.
39:23 Claire Vo? Okay, this has been just so fun. Brandon, thank you for nominating John to come on this show. You're gonna be a hit. John, thank you for sharing your enthusiasm and what you built in your product and your openness to this new technology. Where can we, how can we be helpful to you? Is there anything we can do for you?
39:44 Speaker 00 You can help me fix my VIN number.
39:47 Claire Vo? Okay, we're gonna get this. We're gonna hop off the podcast and we're gonna fix this environment variable. Well, thank you so much. You all have a great day. Thanks for being here.
39:57 Speaker 02 Thanks. Thank you. Bye bye.
39:59 Speaker 00 Bye bye.
40:00 Claire Vo? Thanks so much for watching. If you enjoyed the show, please like and subscribe here on YouTube, or even better, leave us a comment with your thoughts. You can also find this podcast on Apple Podcasts, Spotify, or your favorite podcast app. Please consider leaving us a rating and review, which will help others find the show. You can see all our episodes and learn more about the show at howiaipod.com. See you next time.
Previous A designer's guide to Cursor: How to build interactive prototypes with sound, explore visual styles, and transform data visualizations | Elizabeth Lin Next How to build prototypes that actually look like your product | Colin Matthews (Product leader, AI prototyping instructor at Maven) Page updated November 3, 2026 · transcript and summary are machine-generated