Tech Brew Ride Home - (BNS) Using AI To Manage Your Back Yard

Episode Date: July 6, 2026

This is the article we reference in the conversation: My yard is dying, so I made an app for that (The Verge) Learn more about your ad choices. Visit megaphone.fm/adchoices

This episode is brought to you by Accenture. When your advertising operations fall out of sync, everything else follows. Spotify and Accenture are working together to reinvent the rhythm of ad sales, using automation, analytics, and smarter workflows to simplify campaign delivery and access better data across the business. The result? Less time spent on operations, more time connecting brands with the moments and fandoms that matter most. Learn more at Accenture.com slash Spotify.

Hey, everybody, Brian here. If we have stuck to our schedule, today is our day trip to Mose-Michelle. So here is a HowIAI episode about how one of our podcast friends created their own Android app to manage their backyard. Enjoy. Every day, shareholders meet to discuss important matters about the companies you invest in. Now, you can make your voice heard too. Vanguard investor choice makes it easy to set your proxy voting preference for your eligible Vanguard index funds, whether you hold a Vanguard fund directly or through another brokerage firm, all it takes is a few clicks to select your proxy voting preference and be heard on important shareholder topics like executive pay and director elections. Visit vanguard.com slash investor choice to learn more. It's your shares, it's your voice, it's easy. Vanguard investors own

shares of our index funds and those funds own shares in the companies they invest in. Vanguard Marketing Corporation Distributor. Welcome to another bonus episode of the Tech Brew Ride Home. I'm Brian McCullough. As always, today we are going to talk to someone that I quote from on the pod all the time. Alison Johnson works at The Verge. I should have looked up the amount of times because I can do that now with AI when it can run through my scripts and see how many times I've quoted from you. But Allison, thanks for coming on the show. Oh, thanks for having me. So you posted a piece a couple weeks ago that lines up perfectly with this thesis, which is, hey, I did a funny thing with AI, even though I don't know what I did exactly. So first of all, let's set the table with your not technical. Have you done any coding before or anything like that?

No, not technical. Enough HTML to get me by. Be a little dangerous, yeah. essentially, work in your CMS and stuff like that. Okay, another thing, let me take your temperature on your feelings about AI or your experience with AI, maybe before this project would be a better way to frame it. So like, are you AI skeptical? Have you played around with it before? Just generally, what are your thoughts about AI up until now?

Definitely interested. And I would say a little, you know, more than a little where. I you know my my beat is smartphones for the verge I've been hearing nonstop about AI how it's going to change how we use our phones changed our lives for the past at least three years so it I feel like just now these things are starting to actually ship and there's you know stuff that I'm I've been hearing about for years is finally working on phones I'm using in my hand so Like, it's exciting in that way, and I'm interested in seeing how it develops. But I have had more than my fair share of, like, arguing with Gemini about when my flight leaves or what airport I'm flying out of. So, like, a little tired, I think. You know, I think that I did quote, if you did this piece, maybe a month ago after Google I.O., like the ability to, you know, Google is letting you essentially, hey, with a prompt, you can create an app to be,

to put on your your Android phone. So is that part of what inspired this project? Yeah, definitely. That kind of popped up in the keynote and I was like, oh, that's exciting. You know, there's there's no universe in which I can make an Android app without a lot of handholding and help. So that was really appealing to me. And it kind of got the, you know, got my brain working where I was like, what is something I would like if I could make an app? I think that's part of what all of this is discovering. I've said this before, but Paul Ford was on a, I think, Kafka's podcast,

and he said something along the lines of, well, I have worked with code my whole adult life. So I'm a person that thinks of solving problems with code. and it's sort of like normies like us are discovering that that may or may not be possible. Okay. Enough table setting. Explain what the problem was and then explain how you went down the five-coating rabbit hole. Yeah.

So the problem is my yard. I live in Seattle area. I have a very modest yard that we sort of inherited from the previous owners. of the house we moved into like eight years ago. It's, it's, it's nothing special. There's like some trees and some shrubs and roses and things. And when we moved in, I had no idea. I was like, I, I'd have never touched a pair of gardening shears.

Don't know what to do. So we just kind of like left it on autopilot. That's not great for trees and shrubs as it turns out. So we had kind of come to a place where we were, I was like, I need to fix this. We had all this like landscape rock that was potentially killing all the shrubs. So I was, I kind of had this like growing list of to-dos. And I'm still kind of clueless about how to, you know, I can tell when a plant is sick.

I'm like, there's a lot of flies buzzing around this bush all the time. That seems bad. So I kind of, I needed some help in like an ongoing way and I needed to organize my to-do list. So I was like, you know, why not make an app for that? I'm sure that app exists, honestly. But it was a fun kind of like, ooh, okay, I started thinking about like I could organize, you know, the plants in different zones in the yard. and I could bring in like live weather API and make it kind of like a hub for all the stuff going on. Well, let me interrupt you because, again, this is sort of part of if someone's listening and thinking of going down this rabbit hole,

did you have all of that stuff in mind when you sat down with like your first prompt? Or was it simpler than that? And then as you're prompting, you're layering on things, well, maybe weather and stuff like. How did the process go? So was it fully fleshed out at the beginning or did the process sort of flesh it out as you're doing it? I had kind of a loose collection of things I wanted and not a real solid like framework for how I wanted it to all come together. So I did just like spit a bunch of things out and I used the Google AI studio because that was the one they talked about it. I.O. and you could just get the, um, get in an app onto an Android phone super easily. Um,

so I went in there and I was like blah, blah, blah, blah, blah. Here's all the things I want. Weather. Help me diagnose plants and, um, and just kind of spin it all out and got, um, got a, you know, working preview pretty quickly. And the more I played with it, the more I was like, oh, okay, I, I, I really needed to think about how, um, got a, you know, working preview pretty quickly. And the more I, the more I, the more I, Like, you know, if I'm organizing plants by zones and I have a to do list, do I want those things to tie together in certain ways? So I did have to do some backtracking of like, okay, I got all of this. Yeah, yeah. Yeah, I had to prune my yard app.

Well, let me, right. Or prune your expectation. I think you called it like a list of demands, but. Yeah. Did you experience the thing that, you know, is famous where you say to the AI, hey, I think I want to do this. Is it possible? And the AI is like, oh, yeah, sure.

Like, what was the experience of the back and forth where it's like, did it help you prune things? Did, like, how did the AI shape what eventually you built? Yeah, it was very, I thought there would be a little more back and forth, I guess. I don't, I didn't really know. I just sort of like spit. a bunch of things out and it was like great I'm making a nap right now here it is um so I was like oh okay me you know I think if I had had sat down and kind of like thought through it a little more ahead of time um knowing that it was just going to go off like running uh and make something for me

it would have changed the the output a little bit um I didn't give it any guidance about like what it should look like the color scheme and it came back with just the ugly app I have dark mode yeah yeah right I was like you know let's be a little cheerful let's get some little plant leaves in there and things like that so it was definitely a learning process and I would like to play with more of these you know coding and agent tools just to get a sense of like do are there some that hold your hand a little bit more because I probably could have used it, you know. So also like the sense, okay, let's start this way. So like you said,

you spit out a bunch of things, a list of demands, and you had a preview in five minutes. What's your feeling then? Are you like, oh, this is simple? I didn't realize it's this. Easy done. And then, as you describe in the piece, no, you lost an entire day to it. Yeah. It was a real like, oh, that. That was quick. Yeah. I mean, it hit all the right notes of like, okay, it's got this feature I asked for, this thing and this thing. Great. And I was just like, I fixed the color problem. And then I was like, I'll put it on my phone and I'll just go start using it. And I learned pretty quickly that there was a lot of kind of like testing and troubleshooting that that would have come in handy before. I'd started like. I seem to have lost. Yeah, I mean, the thing it gave me right away, like on the surface level had everything I asked for. I was like, this is great. There was a little tab for, you know, a plant doctor diagnosis and all this stuff. So I was like, cool, I will, I'll just put this on my

phone and I'll go ahead and use it. And I would kind of go out in the yard and start like, documenting my plans and be like, oh, this, this thing doesn't really work. Or there's like, I, I have this way of organizing tasks, but it's not really following the logic or, little things here and there. I was like, oh, this is not really. Let me ask you a question about that. When your company deploys a customer facing AI that misses its mark, who do you think those customers blame? Well, according to the 2026 Delight AI Index, 83, percent of customers survey blamed the brand, not the tech they were using. That's why it's important to have tech you can trust.

Delight.A.I. is a customer-facing AI that delivers hyper-personalized experiences on your behalf. With zero-touch improvement, it can continuously monitor its own performance, find failure patterns before they spread, write the fix, and ship it. It gets smarter with every conversation automatically. The longer it runs, the more your customers can trust it. To learn more, just head to Delight. dot aI slash brew. That's delight. dot a.I. slash brew.

Ever spent hours of your workday tracking down information only to find that it wasn't documented at all? It's an age-old corporate conundrum. Critical knowledge isn't available or accessible to the folks who need it. That's the very problem our sponsor, Scribe, was built to fix. Scribe is a workflow AI platform that automatically turns your workflows into clear-cut documentation. All you do is turn on the Scribe browser extension or desktop app and go

through your processes as you would normally. Scribe builds a guide as you go, capturing every click, step, and screenshot automatically. So what would have taken hours gets done in under a minute. To see what Scribe could look like for your org, head to Scribe.h.com slash ride home. And mention ride home for your first month of Scribe capture free on select plans. That's S-C-R-I-B-E-D-How slash ride home. In retrospect now, was that a U problem or an AI problem in the sense that you got fooled by, oh, this is so simple, but what you're discovering is that, well, I kind of didn't architect, I didn't give it enough to do what I was assuming it was going to do. Yeah, I think it was a me problem, honestly. I think, looking back on it now, I'm like, I didn't have a crystal clear

idea of what I wanted and what I asked for was really like four apps and one um so I think you know the AI kind of did the the best it could with it and came up with us I gave it no structure anything like that and it came up with all of that um so yeah like huge learning from all this was like just go in with the simplest idea and be absolutely sure about what you want and then and then like build from there because it did turn into a whole afternoon and it was nice out too. I was like, oh, this stinks, sitting in front of my computer. Just kind of going back and forth like, okay, well, we need to fix this thing now. And it's like, oh, yeah, sure, got it. And if I'd gone in as like with like one idea. Like I want to diagnose the plants in my yard with and and kind of build on that. I think it would have

been less like iteration and and you know going out and this thing's not working. And every time I updated it, I had to like, you know, load it to the phone again and kind of start over. Right. And one of the things that I've learned with my projects is that number one, I have a greater appreciation for what software engineers do in that sense where it isn't just like, well, here's an idea, one, two, three options go. It is, you kind of have to scope it out and think it through. And if you're saying, oh, I want the, I want the camera to recognize this type of plant, or I want you to fetch weather APIs and stuff, you can say that, but then you actually have to think it through more in terms of, okay, what does that mean to fetch an API and specific APIs and things like that?

So it's almost like doing these sorts of things have trained me to think more in a software way. Yeah. Yeah. It makes me so appreciative of the apps I have on my phone that work really well. And it really, you know, I get that thing where I'm like, do I need to pay for another app or another subscription? and it has kind of turned that around where I'm like, oh my God, there's so much work that is required like on an ongoing basis to make these things even functional. So like, yeah. Right. And you're only building for yourself.

So imagine you're building something for other people where the users would come back to you and say, hey, this isn't working. Right. Right. Yeah. Horrifying. So you describe a scene where like the AI proactively. discovers a bug and is like debugging it for you. And it's saying things like channel is unrecoverably

broken and race conditions and all sorts of things that you don't understand. So talk to me about what you learned about flying blind where like you and I don't know what any of that means. So to what degree do you trust the AI to fix things? And then was that ever like a blocker where it's like, I don't know what you're telling me. So I don't know how to fix what you're telling me. Yeah, it was a really uncomfortable feeling. I think I'm kind of a type A person where I'm like, I want to understand like every, all the context and all the steps of what I'm doing. So just just kind of, yeah, blindly being like, okay, I clicked on the button to fix the bug

and it says it's fixed. Like, great, I'm going to assume that's working. There were a couple of instances kind of like that where like something was not working and I would ask Gemini, you know, like, what do I, what do I do here? What's the problem? And it would be like, oh, you know, that's a really common. That's like, it called something the final boss. It was, you know, in that way that it just is like kind of cutesy and chatty. It was like, that's the final boss of deploying your app or something like that. Like, don't worry about that. And I'm like, do I really, do I take that for face value? And you kind of just have to because I, I,

I'm like, I don't know what the alternative is. Well, another thing, another way to think about this, because again, you go, you and I go out with an idea, we don't really know what we're doing and we're sort of handholding the AI in a way. Sort of like, I describe it as like dealing with a three-year-old where you're like, no, I want you to do this. No, I really want you to do this. No, I want you, you know. But then at the same time, since we don't know what we're doing, where you're describing the, the AI having that sort of person out. Like at the top of the home screen when you started, it says, welcome back a garden mirror. And you didn't write that. The AI writes that. So what I'm

trying to get at is to what degree do you think what you produced was a collaborative thing where there's the AI sort of personality and decisions in it and how much of it was like you? I think it, you know, it did a fair amount of, especially around the structure. I, you know, I wasn't like, I want a tab for this and I want a tab for this. It just kind of built all that. And I think that's, you know, I'm sure the information it's trained on. And it's sort of understanding of like, this is a basic format for this kind of app. You know, it's not reinventing the wheel every time it spit something like this out,

which is, you know, kind of the beauty and the beauty. the drawback of using AI. If you use AI for writing, you know, which I don't do. But, you know, people who are not writers use them for emails or whatever, it's like you'll get the right format and you'll get something that is, if it's the bill, it's not going to invent anything new or creative or exciting. You know, that kind of creativity comes from, you know, a software development. or a writer, you know, thinking things through in a different way. So it's a tool and it's fun. But yeah, I'm like, this is a very like beginner kind of app. And it was also kind of a mess too. Well, I want to come back to what the end result was and if you're still using it. But what I have

encountered, I agree with you that it's not, it's, it's, AI can't be creative on its own. but the degree to which if you understand that the times when it's like oh that's a good idea there's something that i did that recently that added a feature without asking me that i was like oh yeah i should and it just proactively did it because i guess in its mind the thing that i was trying to get it to do it's like well logically it should also like have a thumbnail so people can see the result you know like and i hadn't thought that and So what I'm saying is, right, in the same way that you're talking about writing, the AI can't write the script for you, but like if you probe it and say, give me 30 good ideas, maybe one or two of those are things that you hadn't thought of. And so this is just my own personal thesis, but I'm curious for your take on it where the creativity of the AI is around the edges.

And so that if you go into a project knowing that, then you don't have to take everything. You just only have to take the good bits when it surfaces something that's good. Yeah, totally. There's like, especially in writing, you can get kind of a blind spot or I'm not aware of the phrases I use a lot. And I think there is room for it's helpful to have a tool like that. that can be like, okay, I read your argument, you didn't consider this, or this might be, you're over using this phrase. You know, I don't use a ton of AI in my work that way, but I see the appeal of it where I could be,

you know, before I send a draft off to an editor, kind of be like, am I missing anything here? and that's where I think it really could come in handy for projects like this. Tell me about how this app actually ended up, because I think it was good at certain things like diagnosing this is a plant and this is what the problem is with it. Like how did it end up? Are you still using it? And what are the bits that actually worked in the bits that you're like, well,

that was a fail. It still exists. It's on, I had to switch phones. I had to switch over to iOS for work reasons. And so it's still on like a Motorola razor, but I haven't touched it in a bit. But the thing I did find the most useful was the kind of plant diagnosis. And it was like take a picture of plant the yard, upload it. it would give me a whole kind of diagnosis, but then underneath that, it would have like action

items. Like first you need to do this and then clear out the dead pine needles in the middle of the shrub. And it would add those things to my little to do list in the app. And I was like, that's actually kind of all I need. You know, just distill this stuff because I would have a lot of conversations with Gemini, like, I'm noticing this in the art or what is a good idea, you know, how do I go about doing this thing? And those are kind of like dispersed and I would be, you know, chatting on the phone with it or at my computer typing. It was nice to have that kind of like all in one place and very easily like distilled into like here are the next three things you need to do. Right. Because I think you say at some point, you're like, this could have just

been a Gemini chat, right? Yeah. But again, and, you know, it's a pejorative to say that certain apps or even startups are just wrappers around what the AI does. But at the same time, like, that's kind of useful, like you're describing, if you can take it out in the real world in the moment that the use case matters. Because in the end, that's what software is, is like there's the compute underneath and then just sort of like for the wrapping and the UI and the UX that you put around it so like

again thinking of this like now as like oh this is what software engineers think about all the time your summer travels deserve an upgrade with select Chevrolete, Buick, GMC and Cadillac vehicles connected by OnStar you can stream, game and more on the go

whether you're headed to a music festival or a cottage getaway You can access hundreds of apps on Google Play right from your center display. Explore OnStar Tech Today at OnStar.ca and discover how to take your summer to the next level. OnStar. Better never stops. Was it more useful for you to have this AI out in the real world, even if you could have just ran back in the house and had a Gemini chat? Yeah, I think it was. And I think there is something about, especially with the yard project.

where it's it's a lot of back and forth and it's a lot of I go out in the yard and I do something and I'm cutting back, you know, some weeds and I notice something. Yeah, I'm like, hi, I didn't see that before. This kind of changes things and I would ask about it. There's, it is just a lot of, um, in the moment and kind of like combining my observations and what I'm like literally like seeing and feeling and smell. that, you know, AI can't do and combining it with like, kind of an ongoing dialogue of like not, we're not chatting about everything in the whole wide world, you know, there's something about having that wrapper of like, this is the gardening app.

Right. The focus. Right. Yeah. Yeah. And just being able to feed that stuff in there. And getting out what I needed in a, in a format. that was really helpful because I can't be going through like chat logs, you know,

while I'm outside and it's 80 degrees. And yeah. A few final questions starting with. Were you successful in getting the rhododendron back? Like, what's the state of the yard after this project? It is improving. It's definitely better. I'm like embarrassed at how bad this rhododendron looks.

It looks like Tim Burton's Rototendron's plan right now. But it's coming back. And then there's new leaves on that. We're taking up this horrible landscape fabric and rock. If anybody needs some landscape rock in the greater Seattle area, please get in touch with me. But it's like, then I uncover new problems. I'm like, there's aphids on this rhododendron now. what do I do about that?

So it is, you know, I think gardeners understand this and I'm learning it in real time. And, you know, in software developers too. It's like, it's, you're never done. You just like. Right. Now you need the 2.11 version where you're adding in the aphid and insect section. Yes. Yes.

Okay. These are more philosophical, but from this experience, after doing this, If you had to come down on one side, is it like, oh, crap, we're on the verge of anyone can build software now? Or did this experience kind of exposed for you the limits of that and where we might be far away from that? Yeah. I went in kind of with that feeling of like it was a little bit of conflict where I was like, could I just be paying a subscription for a gardening app that exists? Am I doing some kind of weird shortcut by, you know, prompting my own version of it. And I did come out of it feeling a little more settled with like, I am not going to like Sherlock my way out of subscription apps like anytime soon.

I, you know, I have a handful of paid apps that I love and I rely on. And I'm like that that's not going anywhere. You know, something simple maybe or something like very specific that I want. I had this app that like checks if it's petrorama, which is this like annual like peach event at our fancy grocery store. And I was like, is it petrorama yet? And it tells me yes or no. Like fun stuff like that.

I think I will continue to play with. But I, for one, am just going to pay the subscriptions when it's like something I'm going to need on an ongoing basis that is pretty complex, like taking care of the yard. Definitely. That's not going anywhere. If you had something else, car trouble or, I don't know, we're going to build a back deck or something, you're not going to reach for the AI for one of those projects. at least, you know, hey, we don't know what the next model is going to be like. But basically, this taught you like, I'm not, it's not quite for me yet. Yeah, yeah.

And it's the way I've been using it, you know, because does this replace like we would otherwise just hire a landscaper or someone to, like a professional to come over? It's sort of like the thing that helps fill in the gaps. You know, it's almost like you see, you see a doctor once a year. But then you get little funny things and you're like checking on them. Like, is this something I should talk to the doctor about? And, you know, for better and worse, but I think people are using AI in that kind of like fill in the gaps way. I'm sort of using it in my life of like, okay, I think there's a problem here. I'm trying to understand exactly what it is.

When do I call in a professional? Can I use the right words to describe it when I call the professional? That's where it's been most successful for me. And it definitely hasn't. We still have like a tree guy coming. Like the big trees, I'm not going to let Gemini take those. That's a professional's job. Okay, final one.

Let's assume that someone hears this episode, this July 4th weekend. and does have a deck project or something like that. And they're like, you know what? I'm going to give it a try to. Based on your experience, if someone was going to take a shot, what do you know now that you wish, what would you tell them to do that you didn't do when you went down this rabbit hole?

Yeah, I think like scope is everything. just having, like doing some of that work ahead of time of like, what do I really need out of this? What is the problem I'm solving in recognizing where AI is going to be helpful there in where you need to bring the like, you know, here's how it's structured. Here's how I want things to work with each other. And keeping it simple, I think, is the key to a decent vibe-coded app. But it, you know, it was fun. Like, never in my life would I be like, I made a piece of software, you know.

So it's cool. It's a thing to tinker with, definitely, and can help you out if wielded correctly. Yeah, two things that have come up in just the first few episodes we've done on this is, like, if you work with the AI first to create the scope of work and like the spec, and like say, you know, like in Claude and like you make a markdown file that's like, work with me to make the master spec and then the entire project comes back to that. Yeah. But the other thing that I've found useful is to ask the AI, what am I missing here? Like if you just say, here are my four parameters, it'll just do that. But if you say, here are my four parameters, like, what am I missing? Like, give me a premortem of like where the gaps are, why this would fail and stuff like that.

And then so like what I'm saying is it's not just giving the AI commands. You can ask the AI to help figure it out with you. Yeah. In the same way that like sometimes when I want to design a new feature, I literally say to it, hey, give me a prompt to give back to you that would actually make this feature better than like the three sentences I've given you and stuff like that. Yeah. Yeah. AI keeps, I have to keep reminding myself of that because I think I, with my type A personality, try to get in there and like micromanage and be like, okay, I'm going to tell the AI exactly.

what I need in this scenario. And it's often more useful if you take that step backwards and kind of be like, here's what I'm trying to achieve, you know, rather than going down, like, here's what I want you to do to achieve the thing. And it's surprising how often that is helpful. Yep. Agreed. Well, again, glad your garden is least improving. Maybe this Fourth of July, you'll have some more time to work on it and maybe folks listening will be doing something crazy with AI on their fourth weekend. Thanks, Allison, for coming on

and sharing that story with us. Thanks so much for having me.