Also, check out:
More with Doug:
Contact me. Ask Questions!
- Send me an email here: feedback@doug.show
- Leave a voicemail: (406) 813-0613
Intro and agenda
Hey, what’s going on? Welcome to the Doug Show. I’m Doug Cunnington, and today I’m going to go over, I think three. Three case studies. I’m holding up, uh, random numbers of fingers, three case studies about coding with AI assisted tools, and this is a follow up from last week where I talked about requirements and the general idea of planning for using, uh, AI assisted coding tools.
And it was leaning on my previous experience in. Project management for software companies and that sort of thing. So I got awesome feedback from the email list overall, which I assume it’s people that listen to the show or watch it on YouTube, and it’s a range. So some people have a little more experience with coding.
They had a specific problem they were trying to solve. So there’s a range. There’s a range of, you know, three people.
Voicemail and YouTube slump
And then we’re going to have a voicemail. So somebody sent in a voicemail, which there’s a phone number in the show notes. So wherever you’re consuming this, if you look at the show notes, you could actually like call in on your normal phone or however you talk on the phone.
There’s numbers in there. And then you could leave a voicemail. I think you can leave about three minutes or so and I’ll, I’ll play. One from William today, and I’ll, I’ll just give you an excerpt of it, because some portions were not exactly they, they weren’t irrelevant. But anyway, I’ll just play the part that’s important.
Of course, one thing, it’s not housekeeping exactly, but you know, I assume most people, they replied back, they’re on the email list and that’s why we’re, they replied back. People are downloading the audio version, uh, as normal. The interesting thing is now YouTube has kind of buried many of my videos as far as I can tell, and it’s not, um, that I’m blaming YouTube because I am aware that essentially if people like the videos that they’re shown that more people will see the videos in the algorithm generally.
Is following, following along with like what interests people. If the audience likes the content, then it will show it to more people. That said, you can get you can I guess get unlucky a little bit. Like if a couple things don’t perform as well, you have some recency bias, I believe. We don’t know the exact algorithm, but maybe if I publish a few videos on YouTube and the audience doesn’t like it too much, then I publish one that’s really good.
Maybe that one will not do as well because of the previous videos, because people didn’t like the other stuff that I was doing. All that to say the views are down, uh, dramatically. I’ll come back at the end and mention, you know, one of the symptoms of that. Hopefully I’ll remember to, to mention this. I’ll actually write that down before I forget.
But there is something that I did that could have caused it a little bit more in not, it is not related to like having bad content. Right? It, it’s a little bit different. It okay. I got some.
Case study Annie invoicing app
Got some case studies to go through here, and I’m gonna summarize people gave a lot of detail, which I really appreciated.
Essentially in the email that I sent, I asked people to tell me about their experience using AI assisted coding. So this is from Annie and she says, last year I used chat GBT to help me create an invoicing app for my son’s construction business. What I experienced was anything but straightforward. Maybe things improved over the last 12 months.
I’m not really sure. Commentary for me, I think I did hear. Last fall in 2025, there was sort of a larger improvement where the coding got a much, much better, the coding improved sort of across the board, not just with Claude, but chat, GPT and Gemini. And generally things got better from a coding perspective and a little more straightforward.
Every, everyone that I’m gonna talk about has like their own specific thing that they were trying to do. Thing that they were trying to solve, and it’s great here in Annie’s case that it was a specific business problem. For a real business running versus, you know, if I was trying to come up with some project for me to work on and if I’m, I’m not experiencing a problem where I’m thinking, Hey, I need to solve this.
I would just be making up some arbitrary tool or arbitrary problem to solve, even if there wasn’t a problem there. It’s like coming up with a solution and then looking for a problem that solved kind of thing. So I don’t wanna do that. Of course, there’s probably like example. Coding problems to solve, and we’ve done this, some of us have done this when we went to school and we learned it like programming.
And I took a lot of CS classes and a lot of other computer engineering and coding classes. So I’m familiar. There’s a lot of problems that I could solve, but it’s, you know, not particularly helpful for me to do that. So back to Annie’s. Email here. I was using Google Scripts app, which is a Google customized version of JavaScript to leverage Google sheets, to docs, contact forms and Gmail.
So she was trying to do all of this work on an iPad and work around the limitations that Google places on the Safari OS browser. So right there, that does introduce issues. There’s more complications than is kind of necessary. If someone wasn’t working on an iPad, it doesn’t mean like one shouldn’t do that.
I don’t know, maybe there’s tools that help you along, but just from the platform perspective where you’re thinking, alright, I’m gonna work on a Google suite of products, but I’m gonna do this using safari. On an iPad and you may have issues that you run into ’cause things don’t work exactly the same.
And just having the browser mismatch, safari, the iOS browser versus, you know, all of the Google products that you’re fighting against. So there’s some issues there. We come back to the email. To be honest, the chat GBT coding agent was pure awful. Again, she said this was 12 months ago. At one point I tried swapping to Gemini to get some answers.
Since Google and Gemini are are the same, right? S, which is true, it’s, you know, Gemini’s, Google’s product. But if I watched, uh, Gemini, hand me an answer and then immediately regenerate it and then replace it with slop on one occasion, it erased the whole day’s work. So it ended up not working out, and her son ended up using a paid service, which.
I think they gave like a free trial situation, so that helped out.
Dog interruption and lessons
And Georgie is poking at my legs here. She’s tickling me. Hold on one sec. If anybody has border collie, uh, or a herding dog, a herding breed, they’re just so bossy. Yes. I’m standing here. I stand up when I record. She snuck up behind me quietly and she’s just like poking the back of my leg with her snout.
And, uh, just wants attention, but we’re gonna make it work. So, and he says, be that as it may, I learned more than I expected, but I became jaded by the difficulties and gave up on that particular project to protect my sanity. She mentions that she would try again and have a better idea about planning and mapping things out, which is related to the previous episode.
So overall, Annie was frustrated with the overall process. However, she learned a lot. She probably, you know, made some progress and I am assuming that she understands much better things that are easier or harder to solve. And maybe if she approached a different problem or something a little bit smaller initially, she’ll be able to like, make progress.
The other part, she mentions it here. In the last 12 months, there have been more advances. So maybe things are better now, and if she tried again, maybe it would work a little bit better.
Voicemail from William
So from there, let’s talk to William, who left a voicemail. So I’m gonna just hit play on this.
Hey Doug. My name is, uh, William. I’m from, uh, Pulaski, Virginia. I just, created my first, hundred percent, website. With Claude, I didn’t use Claude Coding, but I just used Claude, uh, directly in the chat interface. And I created an app, web-based app to go, uh, with it. yeah, I.
Planning and, software testing. So I’ll be using some of the tips in there to evaluate, what I’ve been doing. I’m about ready to recruit, the initial cohort of, uh, users for the, uh, app. It’s a, Network, trust-based networking, uh, app. Uh, I won’t go into any more than that at the, uh, at the moment.
But, yeah, your ideas were pretty good I’ll see what I can do to implement them.
thanks William for sending that in and if you want to have your voicemail. On the show, check out the number in the show notes. So in this case, William had a fairly, what sounds like a straightforward sort of problem to solve.
Set up a website. He mentions it’s sort of network based members and that sort of thing. So from. That perspective, it’s maybe a simpler problem, one that has been solved, and that’s a thing that makes it much easier to solve the coding issues out there. And it’s because this is from somewhere else. So it’s, this problem has already been solved.
It’s sort of a canned thing. So I have heard people like, oh, I set up a website, and I mean, that’s. Seems super straightforward. You could set up a website for sure, and it not to discount the advantage of being able to do that just using Claude in the chat aspect. Not even Claude coding or anything more advanced or more complicated.
But he, he sounds like William was able to set up the app, the thing that he was trying to do, and that’s pretty straightforward.
Later in the voicemail, William mentions that he did have some HTML experience, not expert level. I think he was generally familiar, so that. Obviously helped out, and this was a good example of someone with a kind of a straightforward problem in being able to solve it.
Simple problems and stock builds
Now, before we move on to Corey’s example, I, I will mention a similar thing that I’ve seen.
So a friend of mine, David, who I follow on Instagram. He has drone Launch Academy. I think that’s what it’s called, or at least it used to be called that. So he’s running a business. It’s a very successful training business, and he has. You know, a team. He has a lot of processes. There’s a lot of moving parts.
It’s been a few years since I’ve chatted with him, but I’ve seen in the last couple weeks he posted on Instagram where he was kind of doing something similar to William in that he was using these AI coding agents and tools to solve some specific problems in his business. Some were a little more complicated, but.
David’s a smart guy. He was able to go in, figure out a couple things, simple problems most of the time, again, kind of processing data in a lot of ways. However, there was one thing that I saw, he was setting up like a little video game and he, he’s like, yeah, yeah. I created this video game. Like in whatever couple hours this morning I created a video game and it was part of the training.
For his company and he was gonna be able to use it and it’s like a little custom video game. Pretty cool. The thing is, that is probably a kind of a, a stock game situation where like, again, similar to the website, the game has already been designed elsewhere and you’re just putting like. Different graphics and stuff on Space Invaders.
It wasn’t that exactly, but similar to that, where I’m like, oh, this looks like kind of a generic game that is already out there a million times, and it’s just a little bit different. With graphics, it’s a little bit different with the terminology and maybe a couple other things, but generally this is not a custom software solution or anything like that.
Again, not to take. Anything away from William designing his website. Dave was solving problems in his business, and I think from David’s perspective, saving him a lot of time and money and it, it’s actually like highly effective.
Case study Corey AI workflows
Which brings me to Corey. So Corey’s is very successful in Corey’s sort of a serial entrepreneur out there.
And he mentions a couple things, so I’ll read some of his email. Personally speaking, they have been a game changer and he is talking of ai, uh, assisted coding. I started using them in May last year. I’ve used them for all kinds of stuff, things that would’ve been tens of thousands of dollars in months of headache or failure with developers and.
He also mentions, it’s not necessarily easy, but it would’ve been quite difficult. And he says a lot of it has been technical internal processes, and it’s around pushing data from one place to another. So again, pretty technical. I mean, at the end of the day, I imagine it’s, you know, it’s calling. Data from databases and it’s moving it to another place so that it can move through, you know, the sales funnel, for example, or it can move through the rest of the product fulfillment or whatever it is.
Like the details don’t actually matter, and I don’t actually even have the details. He also talked about using AI to help get a website to pass core web vitals, and it was fairly. He says it was very difficult. This was very difficult to do, but it’s something that he is proud of and he spent a lot of time to get it to work his overall point, which I think is fantastic.
I’ll read it verbatim as close as I can to read. Alright. All right.
Planning vs rigid process
The point I wanted to make that inspired the message here, and this is before he listened to the podcast from last week. While planning is important, and I’ve developed my own method for it, a very rigid or over-engineered plan just doesn’t work.
I’ve tried all kinds of methods of having super detailed plans, task lists, and multiple agents to check on other agents and that sort of thing, and ultimately it all sounds good, but the final output just doesn’t work. To build production ready apps and features you need to understand. What and why the AI is doing things and manually check and confirm everything it does.
Corey’s very smart and like I said, he had a successful set of problems that he was able to solve, saving a lot of time and money. Actually, I don’t know if he saved time. He saved money for sure, and he got the solution that he needed exactly. Versus working with developers where sometimes there’s a little communication breakdown, not from a language barrier, but just if someone describes something and someone else is drawing it, then you may end up with something a little bit different because people interpret those words, the same words in a different way.
All that to say, and again, Corey didn’t listen to the episode yet and I didn’t emphasize it, but he’s right, so. It’s important to plan ahead. The plan is important. Coming up with requirements is important. However, you do need to be flexible because essentially one thing that we do know is that your plan is going to be wrong when you’re thinking of, you know.
Fi finances and personal finance or even you know, business finance where you’re trying to plan for the year cash flow. You’re trying to plan projections, you’re trying to predict growth, sales numbers, conversion numbers. If you don’t have historical data, you’re guessing and you have to move forward.
It’s still important to plan that stuff, even if you have historical data for a few years. You still have to deal with the external factors out there, and that can be a bit of a problem because again, the only thing we know is your plan is going to be wrong. So when we’re working with a fin, our financial advisor, I was just on a call with him earlier, one thing we know is like the projections are great to see, but they’re wrong.
We don’t know what. The market’s gonna do. We don’t know what the inflation rate’s gonna do in EV any given year. We can assume over a long period of time, it’ll probably be a certain result, but we don’t actually know what it’ll be year to year. All that to say planning’s important, but, but you have to be flexible as far as being super rigid or over-engineered.
This brings me to. A flashback of software Waterfall method where you do all the requirements and then you do the coding and then you test it, and then you figure out that the users actually don’t like the requirements that we started with at the beginning, and you kind of end up delaying understanding that versus being agile and it’s another.
Version of project management and software development or, or anything that you want to, you know, deliver a project on. But there’s sort of like ongoing feedback along the way where you’re creating things slowly. And I think, you know, it didn’t get confirmation from Corey, but basically he has his own own method for planning and then he’s able to build like one piece at a time, most likely where.
He can take a more agile project management approach and adapt and think, oh, I’m, I’m, I was thinking it was gonna be this kinda software and it was gonna work in this process, but I’m building part one of maybe 10 things that you need. And I see already that the process is gonna be a little bit different than I thought.
So you could adjust that early on versus building all parts one through 10 and then realizing. That you need to start over at step one and fix that first part of the process. So there’s certainly nuance and a plan will almost always be a little bit wrong when you’re planning a schedule. It’s gonna be off.
Things are gonna, most likely takes longer than you think. Occasionally you’re surprised with something that takes less time than you expected, but usually it’s longer. Same with budgeting. Sometimes it’s, you know, a little bit cheaper, but usually it’s more expensive ’cause it takes longer and all this stuff adds up.
Personal projects and next steps
So this has been interesting and I, I have not yet come up with the, you know, a software problem for me, for me to solve where I’m like, okay, I could, I could try something. I did, I did think about looking at sort of maybe a more. Stock example of like something that I could use. Um, maybe there’s already a solution out there that I’m using and then maybe I can figure out how to reverse engineer or come up with my own custom solution or something like that.
But there’s not anything really top of mind where I’m like, oh, I’m actually wasting time on this. It’d be cool to create my own tool to do this. I’ll see if something comes up. It’s been a while since, um, I’ve had like a search on. Software where I’m like, oh yeah, really, it’d be cool if I could do this.
You know, I’m just, I’m not in a spot where I’m thinking about that kind of thing right now.
YouTube ads and algorithm impact
So I promised you I would come back to the YouTube idea and because I made a little note, I actually remembered other, otherwise a hundred percent chance I would, I was not going to remember this. So if you, if you dig the show.
You can, uh, help support it over on the YouTube side. If you’re watching on YouTube, leave a comment. It actually doesn’t matter much at all what you say, as long as you comment. It could be a little bit helpful, and maybe it’ll help the algorithm see that people care about these videos. The thing that I did earlier in the year.
Which may have kind of fucked things up a little bit is I was running ads as a little test for a few of the videos and those videos got more views. That was cool. They were getting like a lot more views and it was one of those things that I did back in the day to help, you know, boost my subscriber count.
I was like, ah, I’m curious, like what happens if I run these ads? So. As a result, I also got a lot more subscribers. So there’s like new subscribers out there. I don’t think they’re like high value, highly interested subscribers. You know, I think they were shown ads. They watch, uh, some of the video they, and it makes my subscriber can’t coop.
Fun little game. I play super cheap. And the thing is, I’m pretty sure those people. Don’t care about the videos. They don’t care about me or the content or anything. I don’t know how it works at the end of the ads or whatever, but these people are just watching the ads and they’re not watching any other shit at all.
So one thing that could be happening is many of the recent videos are maybe suggested to the people that don’t really care about the content because they were recently shown an ad and interacted with it. So they’re seeing some of my other content and then they have no interest in it, and then that’s hurting the overall channel.
So the, the numbers are even worse because maybe. I mean that, that kind of makes sense to me from a, a logical standpoint. So we’ll see how it goes. I have thought about you know, coming up with some sexier titles o over the one, like from last week, which was horrible requirements for AI assisted coding, boring, right?
I mean, someone might search for that. And then I think. Actually, I don’t know many people would be searching for that, but you know, it’s not really driven many views and it was actually some of the worst views that I’ve seen in the last, actually, worst than ever. Even when I first started, I got more views than that, which is kind of crazy.
So we’ll see how it goes. I, as I mentioned before. I know the algorithm is really based on, like, if people are watching the content and they like it, the video’s gonna be shown to more people. On my other Channel Mile High Fi, ironically, it’s blowing up. I, I have a couple videos that are, you know, some of the best videos that I’ve ever released on the channel as far as like getting a large number of views in a short time.
It’s, uh, it’s fantastic and I’m like, oh. Yeah, it, it can work out. So it’s kind of crazy. That channel is growing a lot, getting a ton of views, a lot of interaction, the things that YouTube does wanna see on a video. And then this channel is, you know, as low as it’s ever been. But I’m pretty sure if I publish something better that people actually like it.
It doesn’t need to align with the stuff that I wanna do. I have a couple ideas. I have, um, one where I actually have a decent script and it would have B roll. It would not really be a podcast episode, but it would be an actual like YouTube video created intentionally for YouTube. I stopped drinking caffeine for about one month, little over a month.
And I often, uh, consume a lot of caffeine. So it was an interesting process and it actually worked out pretty well. I’m back on the sauce though. I do enjoy caffeine and I am, I’m having less, but I am back on it and I do enjoy it. It is so good. I, I mean, I love coffee. I even like those nasty monster energy drinks.
I get the stuff with zero sugar though, so.
Wrap up email list and outro
Alright, I think that’s it for today. If you. Want to be on the email list, which I highly recommend, follow the link in the description. That’s something I should mention more often. And the song that Welcome Dust and the one that I’m gonna play here is one that I played with, uh, Matt Jovani.
He asked me to, uh, assist with, I think this is the one. Or did I record this on my own? This is, uh, one I recorded on a resonator guitar, and actually it, it’s Matt Japanese’s tune, but like, whatever I played sort of fit in a little bit and he put it on one of his videos for Brew Cabin out there, and I forgot that I, this, so right here.