S13 E3: No Code? No Problem (with Billy Howell, Founder of Stupid Simple Apps)

20:34Nik:Yeah, totally. Yeah. So like Cursor, if you want to be really custom, you're doing something that's kind of atypical or has lots of moving parts. Maybe you want to go slower even. Replit or lovable if you want to get to like a quick prototype, proof of concept, something super easy, like a database with username and login. but you can build way more complex apps in there and I do. Right. And so how do you use ChatGPT versus going into a cursor or Lovable or Replit? Yeah. So I just go in there and I use this app called Super Whisper which I need to get a referral code too because I keep plugging it in all these calls but it's super cool.

21:11Yeah. I have the Replit one. Throw that in the bio. I will. With Super Whisper it's voice to text and I think it's, I'm still on their free plan. And so I find that when you're talking to AI, it's more helpful to be verbose, right? Like super helpful to be verbose. And you're just limited by how fast you can type. So I go into chat GPT and I just describe the app like I would describe it to you here. And then if I have kind of unknown things, like if I'm not sure how the app should look or how we should do something, I'll ask it to do that research. And then at the end, I'll ask it for what's called a PRD.

21:48to product requirements document, which is kind of something I've stolen from like software, like real software engineers, right? Not computers. And I just list like, here's all the features, here's all the screens, like the pages we need. And here's like the integrations we're going to use. And then I'll just take a look at that and pop it into like Replit or Cursor and tell it to get to work. I find that makes your starting point a lot better because if you start out incorrectly out the gate, and I think this is where a lot of people get stuck, right? They've got an idea that's like, and they put in a sentence or two, and then the idea doesn't look anything like what they pictured in their head.

22:29So they're like trying to unwind all this code, you know, hundreds of lines of code. So you want to spend two minutes to get closer to a fully fleshed out idea before you start these. It'll save you two hours of time. Yeah.

22:45Nik:Yeah. So two things I picked up since we started, like two things that matter a lot. One is organization of data. And the second one is the prompting, like the inputs you give it. You know, have you figured out any tricks to make prompting better, smarter, more accurate? Like I was even scrolling Twitter yesterday and I'm sure you saw like the videos of the Ikea, the VO3 IKEA ad. Did you see that? Where like out of a box, it just all opens up and everything animates perfectly. And the way that it's written, like the prompt going into VO almost looks like code. It's like camera shot equals this, this height, this, this.

23:29Nik:So I'm wondering, are there tools to prompt better or are there tricks to prompt better? Because most people, they tend to assume, and this happens in marketing too, this is why so many brands don't take off from a marketing standpoint, but they assume because they're drinking the Kool-Aid, whoever they're speaking to also understands the same level of context. And I don't think like chat GPT does a really good job asking questions unless you tell it to ask you questions. So have you figured out any tricks or are there tools that exist to do better prompting? There probably are tools. I don't use them because there's just too many.

24:07I don't have the time to go experiment with them. You can get good enough on your own. So one of the things I do, it's funny that you say about questions, is in my system problem in ChatGPT, I ask it to ask me three clarifying questions every time I send a prompt. And that just really helps get it thinking, get me thinking. So that's like number one, I would do that. And then whenever I'm coding with like an AI, like coding agent, and I'm working on adding a new feature or fixing something, I tell it to, I'll explain the problem. I'll say, diagnose the problem and find the solution. Then walk me through it, but don't write any code.

24:45Because if you think about it, when you send a request, ChatGPT has got, say it's got 100 % of brain bandwidth. It's going to spend 40 % on thinking about the problem, then 60 % on coding, because it's going to think and then code. But if you just tell it to walk me through the problem, it's going to use 100 % of that bandwidth to find a solution for you. It's going to poke around more because it's not thinking about saving tokens. So I found that really helps.

25:12Nik:Oh, that's really interesting. So the loophole there is you're basically telling it to conserve tokens because it itself is trying to be efficient on the output. But yeah, and I even will tell it if I know something is going to be like a multi-step thing, like it's going to build something out to come back to me to build more. I'm going to say, don't worry about tokens. Like we can do this in multi-steps. I don't know enough to know if that makes a difference, but I suspect it does. And the other neat thing about, let's see, Windsurf is kind of like Cursor. I think Cursor even has this now, but you can toggle in the chat, in the input rather, whether you want to chat or edit code.

25:56So you can just cut it off from coding, which is really useful. And I recommend doing that. And then the other big thing is if something breaks after the agent has coded something, don't add more code on top of it. Roll it back. Most of these AI coding softwares have the ability to roll back to a checkpoint. So roll it back and then try again. Don't keep putting good code on top of bad code. Being disciplined about that makes a really big difference.

26:28Nik:Yeah. How easy is it to now take, let's say you were able to build this web app and now you have a dev team at your brand, for example, and they want to go and modify the design or the CSS or whatever it is. How easy is that? design in css is really easy like in replit in most of them you can just like even click like a little pin button and then select elements on the page you can manually edit it in the chat like it'll show you all the css values or you can say like delete it or modify it in this way or do this uh so non-technical people can very easily edit the design if you just like edit uh add them to a repo so that's a good question no one's ever asked me that before yeah because i always think Like a lot of times people will think, well, you know, let's say I make an app, you know, is it easy to continue innovating or a year later, let's say somebody new joins the marketing team and I'm gone, you know, are they going to be able to go and take that from where it is and innovate?

27:24Nik:Or do they have to go back into the, you know, the text, the chat interface to make it to that, whatever it may be. Yeah. Yeah. I might even be able to just show you one. It's, it's really like, you can literally just like you're in like a bubble or web flow, So just click in into the preview, in Replit at least, and change it. So that's a good point because, yeah, you're going to have a lot of different teammates working on these sites. Totally. Another good practice is to separate the app from the actual landing page or like the marketing pages. So at the beginning of the prompt, tell it, like specifically say, don't build a landing page.

28:06This is just the app. And then you can deploy that to app.your subdomain. And then you can make a landing page either separately in a different AI or just in Webflow or ClickFunnels or whatever you want to use. I feel like probably people might not think of that, but that's definitely something you could do.

28:25Nik:No, totally. Makes a lot of sense. What are some scrappy use cases you've seen being built very quickly and adding a ton of value? Like lowest investment required for a pretty high value output. Probably what I'm doing, right? Like B2B. Finding people with hot pain point problems at their business and then building them a custom solution rapidly, you know, three to four weeks, right? I think that's the quickest path to like getting outsized returns on Vibe coding versus making a D2C app and then having to figure out the marketing and the distribution, right? So I'm always going to preach to go B2B first, especially if you've never done either because the average deal size is higher.

29:22It's more relationship-based, one-on-one versus D2C where you might have like 50 customers, but they're all paying you$10. And it's hard to separate signal from noise on what to build. And so I think, for example, like AI receptionists for local businesses, building those custom or even just setting those up from an existing product is a big opportunity. Chat bots that are trained on companies' data, Like we had a guy who manages like a restaurant group that does this charity thing every week. And he wanted his menus digitized so that people can go on the site and be like, I'm in Fort Lauderdale on Thursday and I want steak.

30:07Like, which one can I go to for this charity event? So, yeah, little custom AI bots for people's data. Definitely useful. And that's big in e-com too, right? Shopping assistance. Mm-hmm.

30:23Nik:Have you started to see an internal full-time role of AI coders or optimizers, or I'm not really sure what the title would be, but somebody who's going into a company and saying, wait a second, we've got all these processes that are being manually done. We can start turning these into apps or web apps to really save time or just make things a lot more efficient. I don't know that I'm aware of that being like an internal role anywhere, but I think you certainly could make the case for yourself at your job and be that person. Because it just doesn't take that much time to spin up something that's like visually impressive and like demonstrates the value.

31:08and you know since I've done a bunch of these podcast interviews people DM me and be like hey I sold these like AI coding solutions to my boss like to my CEO for 20k 30k right and their approach was basically build it out some of them built out the whole thing I would build out just like a demo and then take a loom and send it to the CEO or set a meeting with the CEO and demo it for them there and see if you can get buy-in. And even if they don't like that particular solution, they know, all right, I've got Billy in my back pocket now. And they'll come back to you when they see something in a couple of weeks that they think could be automated.

31:47So yeah, I used to do this when I worked in consulting was try to automate stuff in Excel when we were doing repetitive data entry. And it got me on some cool projects and meet some cool people in the company I wouldn't have. So I think it's a great idea.

32:02Nik:Yeah, I feel like in the next like two to five years, there's going to be there will definitely be if not an AI team, I think a lot of brands or a lot of companies are going to have some sort of like AI, you know, consultant or a full time, you know, one person who's constantly trying to figure out how to make things more efficient. Um, you know, I even remember like when notion started to become really big, this whole crowd of notion consultants and, uh, and whatnot became really big because it was like, wait a second, we can make information a lot more efficient within the company. And I think that's going to happen here.

32:36Yeah. And it's going to get to a point where it's so accessible that you're going to need an expert to point to that can say no to everyone being like, we should automate the coffee machine, you know? Yeah.

32:48Nik:When I wake up, it should know that 37 minutes later, I should be at my coffee machine. Yeah. Okay, cool. I'm curious. So like, with all the apps that exist today, Shopify, Klaviyo, you know, PostScript for SMS, or Triple Whale, like, a lot of these apps are starting to build AI. Where do you see like the path? You know, the way I kind of see it is, I feel like a lot of these apps are going to try to build AI or leverage APIs of different models within their apps. And then as that becomes more familiar, we're going to start to see like marketplaces where I can go and download, you know, Billy's apps and just use them in my own store, or I can subscribe to Billy's bot or something of that nature.

33:35Nik:How do you kind of playing out? Yeah. I think to me, the interesting thing is most of these products use these generalized LLMs, right? Because they're like pretty good. And I think we'll see, I think we'll see a gold rush maybe of truly custom LLMs that are really actually trained on your data or data from your industry, right? That's making them really good at understanding like, Like, you know, I don't know, skews for like fashion rental companies, like pricing strategies for painting contractors. Right. I think that would be a huge value add. And I think it's something that I think it's something that people aren't focusing on now, like at all.

34:18Right. And I think that's kind of where the AI stuff started was everyone's going to have their own AI that's trained on their subject matter. And somehow we've just kind of blown past that. And we're using the, cause the generalized ones are so good and they do, they're trained on like the whole internet, uh, theoretically. Um, but I, I think we're going to move towards like customized, like this is an LLM that's trained on like my data and people in my industry. And it's really good at doing like these four things I needed to do.

34:46Nik:Yeah. I even think like, you know, if there's one guy who is such an amazing media buyer, like why wouldn't there be a bot that he could or some AI or app that he has that does a weekly audit of your media buying or a, you know, weekly website audit just takes all of this person's learnings and data points and whatnot. or to your point about the inventory thing, you know, if somebody managed pre-order inventory at Rent the Runway for years and understands how to do that at scale, you know, somebody who's running, like Ben at Latiko could just be like, wait, I'm just going to subscribe to this guy's app now and use all of his learnings to make sure that mine is perfect.

35:24Nik:That would be really cool. I'm like picturing, like, you do a video call once a week to just like brain drain yourself and it asks you smart questions to train and then it comes back, It tries to solve problems. It comes back. And then you can just kind of train it passively over the course of like a year. And you'd have a really well-trained bot. Is there any sort of like, how have you best seen a way to like download your brain?

35:56I think people are doing cool stuff with like Notion or like Obsidian, like taking in all their apps. um i use granola lately which like records all my calls and then i can search in there but like as far as my brain i would have to take in like all my video content all my written content all my emails and then synthesize that somehow synthesize synthesize um i i don't know have you

36:24Nik:seen something in this that's useful no i i haven't but um you know i've done i've done the thing where you try to upload as much, you know, as many podcast transcripts or newsletters or that kind of stuff. But like, to your point of the models, you know, like the whatever model it is that you upload into, it does not understand tonality, like the same way that if you were to voice record in chat GPT, that understands tonality, which is really interesting. So like, So there are definitely ways to upload and then search, but it's still, at least from what I've seen, it's still not like, for example, we have this thing called NickBot in our Sharma brand Slack.

37:08Nik:And it's basically got everything that I've put out on the internet uploaded into it. And it's pretty solid. And we've done a lot of work to try to make it more and more accurate, but it's still not near perfect, I would say. Yeah. Because it's got to cheat, right? It's got to run SQL queries in the back end to find, oh, he's asking about this podcast. Let's search the transcript. It can't. The context window is like, it just shows you how complex the human brain is, right? These things are really smart. They're not anywhere close to what our brains can do. Yeah. Well, our brains too, I think, are like the more I use AI or the more I see people using AI, whether successfully or unsuccessfully.

37:51Nik:The real kind of secret sauce is in knowing where to turn and navigate and where to go next versus just grabbing the answer and then figuring out what to do with it. Like, you know, okay, if this doesn't work, then the AI is not going to know what to do next. But if you know what to do next, like that's kind of that secret sauce. Yeah, absolutely. Think of them as like a really smart, like high school intern or something. Yeah. Yeah. I feel like the notion that AI is going to make people dumber is, it could be argued against pretty heavily the same way that like Wikipedia would. Yeah. I think it's going to make people crazier though.

38:30Nik:For sure. For sure. The way it kind of just affirms. Yeah. Impatient is right. Every time I build out one of these apps and I'm doing it with my buddies and I'm like, oh man, it's taking like five minutes to spin up this app. And it's like a miracle. Like a year ago or two years, whatever, I'm copying and pasting stuff one file at a time into ChatGPT. And now we're getting impatient. That's funny. No, totally. Or even ChatGPT just unlocked agents. And I was running an agent yesterday and it took 13 minutes. And I was like, are you kidding me? This is what I'm paying for? 13 minutes? I was going to say that on the brain thing, I saw a good tweet today that was like, why isn't there a ChatGPT?

39:13for my life where I can just be like, what does my neighbor do for work again? Yeah. Stuff like that. I'm kind of working on like a solution for someone that pulls in from their Google calendar and their email. How, like, when's the last time I was at an event or had a meeting with them? How many times do I email them this month? And it gives them like a weighted score. I'm like, this person's really hot or this person's colder. I think that is, it's not a way to download your, your brain per se, but I think that's a really useful use case, just knowledge base of your friends and your family or, yeah.

39:48Nik:Yeah, I agree. I was even, I was thinking last week, I was abroad and I met so many cool people, you know, anybody from like a security person at a venue to a bartender to somebody who's also vacationing. And, you know, I probably collected like 30, 40 Instagrams, but the problem is like, I don't remember where I met them or in what context, or, you know, I'm sure I took a selfie with half of them to remember, but those were not attached to their Instagram or their WhatsApps or, uh, and whatnot. And, um, yeah, so many use cases like that, that I think are just going to be instantly solved. That, that, I mean, same thing that I always forget to do the selfie thing.

40:26It's so smart, but I was, I was digging for this guy's number that I wanted to follow up with. Um, I was looking on like LinkedIn and stuff to find him. I had his number on my phone. He called me like the next week and i was like i it's just a first name and last name it wasn't connected to like your business or anything so um some sort of ai yeah like personal assistant like that would be great it would be crazy i remember back in the day there were so many of these like um ai uh you know

40:58Nik:calendar scheduling assistants that would that were basically like if then yeah scripts um but I can't imagine like what that next generation is going to be with the ability to now plug all these different models in. The main thing I think is for stuff like this, like it needs to be like in the primitive apps that you use, like Google Calendar, iMessage, Telegram, whatever. Anytime you've got to download a separate app or like open a separate tab, like it's just making it kind of more of a pain. Like it needs to work with the primitives of like your OS, like your contacts thing. Um, so yeah, come on, Apple, open all that up more.

41:39I know it's really like, uh, Apple is really a gatekeeper of that.

41:44Nik:I mean, even like I'm, I use iMessage as a, as a, as like a business app constantly. Like I need iMessage pro, but that's never going to come with Apple. Yeah. Yeah. That's a good point. Yeah. Yeah. Okay, Billy, first of all, thank you for coming on and just like filling my brain with excitement about AI. I feel like now I want to run through a wall and build some apps. Yeah. But where can people, like people want to hit you up if they want to hire you to build an app, if they want to, you know, jump on a call like Troy did and just consult with you for an hour about what they can do, how do they find you and how do they get in touch with you?

42:25Yeah. Stupid Simple Apps is my agency. So if you Google it, stupid hyphen, it's got hyphens in there, stupidsimpleapps.com. You could fill out a form for us to build your app. On Twitter, I'm Billy J Howell. And I've got a link in there to like accountly if you want to do like a one-on-one strategy call. But yeah, go watch my YouTube channel too. It's the same app. I've got kind of my strategy for building and selling your first apps in there. It's really approachable and you can do it. Like I promise, just people don't follow through, right? So like you have to be the one that does it. But there's a huge, we didn't even talk about like Upwork, but there's just like people sitting there ready to pay you money to build custom software.

43:07And I just think that's so crazy and exciting.

43:10Nik:Totally. Yeah, I will say your content online is insanely tactical and it's very much like somebody can look at it on the left screen and they can execute on the right screen. So I highly advise that you guys check it out. And if you have any questions, like the call you book with Billy is going to be the highest ROI call of your life. So highly recommend doing that too. Thanks, thanks. Yeah, everyone super happy from it. And I just want to help people execute on this stuff and make like a plan to go build and like sell apps because I promise you can do it. Like I'm not that smart and like I figured it out.

43:44You just have to execute on it in practice. I mean, that's the crooks of where we live with how bountiful the internet and software is now is you can build anything. You got to just have someone tell you, hey, dum-dum, build this and then build this, right? It's not going to be a hit right away, but you're going to learn from this and then you'll get towards whichever one makes you money, right?

44:07Nik:Totally. And the seventh thing they build is going to be amazing. Yeah, on the seventh day. Yeah, amazing. Cool. Well, Billy, thank you so much for jumping on and I'm sure we're going to have you back soon. Awesome, man. Good to talk to you, Nick. Thanks for listening. We'll be back next time to cut through the noise on CPG, retail, and e-commerce. If you enjoyed this episode, why not share it with a friend? And be sure to subscribe wherever you listen so you don't miss the next one.