I Built An App With Claude 3.5 In Less Than 10 Hours

On today's show, we are giving you the complete tutorial of Claude Apps, one of the best AI features to help you build anything that comes in your mind as a web app. Let's get into today's show. Kieran, to say you are Claude Apps fan might be an understatement. I feel like you're a cloud app super fan.

I am a cloud apps. I'm a Claude fan super fan in general. I think the feature they've added, which just allows you to create the app but then post the app to Claude is like one of the most coolest features that any of these LLM companies have launched because now you can create lightweight apps, click publish and immediately have something shared on the web. So that is the big change that Claude have made made over the past week. We're going to show some examples. I'm going to get into a quick example of how I've used it to create a minimal, viable version of something and maybe just talk about what this means for all of our listeners that this functionality now exists in the world.

Do we want to maybe start off by just showing some examples of what we're talking about? Because I know people out there watching, a lot of folks probably haven't seen or don't even know that Claude Apps are a thing. That's why they're watching the show today. So maybe let's show them what we mean.

Yeah, let's show them what we mean. So, and you know, in days gone by where I was showing someone, I was showing folks how to do this, right. In a previous episode, I kind of showed this kind of cool thing you could do to get Claude to write better, which is basically I can combine a post template with a write in template to actually create some really cool content. Now in the world we have lived in prior to this functionality existing, that's where my kind of journey with this would have stopped. Right. I sent it to my newsletter, we did a podcast episode on it, but it was kind of like a tutorial, like a, you know, kind of tutorial that anyone can give where you say, hey, there's a document. There's a document. Combine the documents and write this prompt. And because. Because I can't really code that well, right? Even though I should be able to, because I spent a lot of time.

You did go to university for coding.

Yeah, I did go to university for coding. Let you off the hook, gloss over that part.

That's where clearly you studied hard.

And so I would have just stopped, right? And maybe I would have, maybe I would have released some sort of like course, email course or something like that. Well, what's really exciting today is if you have really good ideas and actually that's a pretty good like way to get signal is to do it through content first, then you can actually progress to, to an app. Right. So in that example, when I sent it to my newsletter, I immediately got tons of people coming back to me and saying, hey, do you have more post templates? Do you have more writing styles? Do you have ways that you can just launch this as a product? And so I went back back into Claude and I was like, okay, well, let's see what I can do now. This is a very minimal, viable version in that there's not. I actually, I actually in the next week will have multiple post templates. I do have them, I just haven't uploaded them into Claude. And so what I mean by post templates, if we kind of, again, just to. If you have not listened to the previous episode, let me just show you what we mean by these.

Yeah, can you maybe show how you get to this like, fancy app that you're showing here? So I think so that people understand how you get from the kind of Claude Basic, you and the AI going back and forth to like, hey, I've got this fancy panel here that looks like it's doing some real things.

So what we start with, we're going to take one platform. We're going to start with LinkedIn, right? We kind of did this last time. And so what you would suggest with any of these platforms is there's these kind of formulaic templates that work really well. And so we kind of went through this. If you, if it's your first time listening, you should go back, listen to the episode where we taught people how to use these templates. But you can see basically there are a series of different post type templates, right? There's one called the Personal Achievement Story, and we lay out exactly how to recreate that style of post. There's one called the Industry Insight Revelation and we talk about exactly how you can create that post. There's one called the Problem Solving Showcase. Right. And again, we talk about how you can exactly solve that post. And there's one called a Mentorship Magic Story, which is basically, you know, engagement. Uh, talk about a pivotal moment in your career when you felt stuck. Introduce introduction of a mentor figure with vivid details that bring them to life. Like again, it talks about, it gives exact instructions for the AI to be able to replicate that. And so when you think about what we're doing in an app. Right. We, what we can now do is tell Claude to basically create an app that allows someone to pick one of those templates. You see, I've only got them one. I've only got one loaded in, but I will have multiple of these loaded in. Then we have write in styles. So I kind of showed this on the previous episode. I'll just give you one quick example here.

Well, he's pulling that up. If you haven't watched the previous episode from like two weeks ago, you should go and check it out. It's really good. Hit that like and subscribe button too. This is, this is pretty, pretty gold. Giving you a nice walkthrough of really how you can take your AI interactions and actually start to make them closer to like a full fledged web app. Which is pretty awesome.

Yeah. So now you have a writing style, right? This is Gary V. Gary Vee doing pretty well. Got about 5 to 600 likes last week from some Gary Vee posts. Which means basically I created them in this style guide and then I heavily edit them for all of the people about to jump on. Oh, you're copy and paste and. No, I heavily edit them. What I've actually started to do again, I'm going to get back into the app, but what I've actually started to do is actually not ask it to. This is a, you know, fresh off the Kieran hot desk of using Claude that just happened today. I've started to ask it to give me talking points, not the post. Basically give me Gary V phrases.

Oh, I like that.

I had a. I had a post today that's doing pretty well on LinkedIn and it didn't write the post because I still do think that I have to heavily edit it to make it really good. But I've realized that I really good at kind of Gary V. Fraser. So I had a good one. Like hire talkers. Hire doers, not talkers. Don't build a team of powerpointers. So I basically ask it for like seven phrases and then I try to figure out if I can integrate them into the post. But anyway, this is kind of a writing style, right? So basically it details exactly how Gary Vee writes. So let's get back into the app again. My journey would stop by creating those things and then kind of saying, oh well, I'll just teach people how to do it. And I'm like, no, no longer. Because now I'm a developer so I can pick the writing style, I can pick the post type and Then I can give a post topic. So more, more craft, less PowerPoint and marketing. Now what this app should do, like I picked the template, pick Sean Perry, who is a really great writer. It should basically generate the output. So why have I got generate prompt instead of generate output? What Claude cannot do yet. You can't actually ask Claude to do this prompt, right? What actually I want to do is to say, okay, send this to Claude via a prompt and then Claude send back the output to this app. But because I cannot. It cannot call APIs yet, but I think they will add their. At least their own API because it would drive up usage of their own API. Right?

Huge retention.

It will generate the prompt. Right. So again, this is minimal viable version. It is not the best version I could do, but it will say, like create the LinkedIn post on this topic. Then it will say, you know, it's picked the right post type and it said, here's the exact template to copy. Make sure that we're doing it for our listeners. What you're seeing right now is an app live on Claude where the prompt basically details the exact post you want to create. Then it gives an exact template for the post style on LinkedIn. And then it basically gives an exact write in style for Sean Perry to create that post in that writing style. Now what the user can do now is just copy this to Clipboard board and go run it themselves in Claude. I suspect what you they'll be able to do in the future is just actually go get the output from Claude itself within the app. And here's like. So let's. How good is this? What do you see this? Right. What would be kind of cool here is I want to show you how quick it is. We want the user to be able to pick two writing status. We want to ab test things. Right?

Yep.

So let's, let's make a quick update to this app. We want the user to be able to select two write in styles to create two competing prompts so they can a b test the post. Could you recreate the app with that new functionality? Now I always have to remind people how mind blowing this is. I think we're just getting too used to it. I'm just like literally coding here. Updating. I'm updating functionality. Right. I'm basically creating new functionality for an app. And I wrote three sentences. Right. This is Kieran back in his college days figuring out how to reverse engine.

If college Kieran had figured this out, you wouldn't be alive today. You would have had all your Time to party. It would be bad. It would have been bad.

College Kieran tried to actually figure out if he could copy his project at the time. Because I was like on. I'm such a. I was a growth hacker before growth hacking was cool. I find all of the ways. Now look what it's done, right? If you're looking at this screen on YouTube, what's it's added? I can now create. I can now select two write in styles, right? So I can basically say I want to test Gary Vee against Naval. And then I can put less PowerPoint, more craft in marketing, and I can generate the prompts. So now look at this.

Well, what's.

What's really.

What's really interesting exactly is that not only is it writing great code, but it's understanding what you want. It even. It's even labeling them A and B because you told it was you were going to do an A B test, right?

It's going to do an A B test.

So it's not like prompt one and prompt two. It's like prompt A and prompt B. We're writing an A B test. That's what we have to name them. And then it's pretty amazing.

I don't know how you can't get incredibly addicted to this. Like, I found it really addictive. I was like, all night, I was like, I need to stop doing this because I'm getting a headache. But it's just like so addictive. Like, I'm just literally building code now for the real app. I'm actually going to build it and deploy it and repl it, because I still want to call that I want to get the output and deploy the API.

Hey, everyone. From one marketer to another, I want to help you out. It's tough keeping up with the content demands each year, especially across a variety of content channels. Balancing the workload while still maintaining quality can be a real challenge for everyone. We've got a free guide from HubSpot on how to use AI to make that all easier. It walks you through how to integrate AI tools into your content workflows so that you can come up with more ideas and spend less time implementing them. This is going to save you a ton of time and a ton of money. I don't know what else I have to say. I've dropped the link in the description below. You're going to want to go check it out. Now, back to today's show.

We also have a special episode. I give you a little preview of that. Our incredible founder co founder of HubSpot Dharmesh is building something really cool and he's actually building this functionality into the platform that he is building. And I think we will be able to do an episode just on that. And actually when we do that episode, it will be available for everyone. And it's got some extra cool functionality that I've been working with him on that he's integrated that I think is even cooler than this.

Sweet.

So you can only get that if you follow the show, hit subscribe, RSS, hit subscribe on YouTube and maybe Dharmesh will come on and talk about that a little bit. But we'll definitely do a quick showcase of that that he's building really for our listeners because it's based upon our workflows. So let's get, we can get into some other examples, but it is hard to like figure out what is going to happen here. Right. Because this is the point I made. What we made with Andrew. Andrew Chen was on Amazing conversation. You should go back and listen to that. But let's say that there is a hundred people who want this app and I can charge $10 a month.

Yes.

It's now economically worth me doing that. Yeah. Because it takes me so little amount of time to do the app. It's just the idea. Right. The idea has to be really good. But the app itself is such. Yeah, let's.

Hold on, let's. Let's do the math for people real quick. Okay, but by the time you said that you're gonna publish a version of this on replit, right. Because you just want to learn how to do the thing soup to nuts, right.

The API.

And yeah, you wanted to have access to Claude API. So instead of generating the prompt that the process go, put it in Claude, it generates that work output that everybody wants from this all in. All in how many hours you think you will have in on that by the time you hit publish and like the apps live?

So I would say less than 5 to get the app done. It did take some time to really figure out the write in the post templates in the write in style.

So call it 50 hours.

No, no, 10.

You're gonna call it 10, maybe less like eight. Let's say eight to 10 hours all in. Let's say we'll just. Let's sit for this math, say 10 hours. And let's. And so basically if you get hundred people to pay you $10, that's $1,000. Right. And so right off the bat, you're making a hundred dollars an hour in Month one in month one, right? Month one, month one of publishing this hundred dollars an hour. Most people I know would love, love to make $100 an hour, right? And if you can keep half of those people paying you a second month and, and get another 10, then you're going to make $1,500 and you're going to make $150 an hour, right. And you could see if you actually worked on something valuable. A lot of these things that take this amount of time aren't going to be that valuable and it'll be quick hits. But like you find a couple of these things that are valuable, that have a year shelf life, you could easily be making tens of thousands of dollars for ostensibly like 10 to 10 to 20 hours with work because you still would have to have a lightweight website and you still have to do a little bit of social promo. You got to have some billing. So you're, but you're, you're probably talking post launch maintenance of like minutes a month, not hours a month, right?

And it's how quickly I can add features. So if I'm like in month one, this feature, a user has said, hey, have you actually add this feature? It's actually much more valuable. So let's say they say, hey, we can really, we would love to use this for newsletters. I can literally just add in this an option then to start to do it. For newsletters if they said, we would love to be able to upload a post that we want to do and then apply the writing style and the template to that post. So we have a version of the content we want to post, we want to upload that and apply it to that. I can probably get that done in like a couple of sentences, right? So the amount how quickly I can actually start to make the product more val, how quickly you can integrate user feedback. Yes, I can, I can basically, you know, you could literally start a product suggestion form, have the LLM agent actually ingest the suggestions and create prompts for them and just do that itself. Like I would need to be involved, would I? Like I just have the users on a forum saying here, can we have these features. The AI agent has a feed of that. The, then it passes it to create the prompts and the prompts ask Claude to add the features.

And. And these were just going down a rabbit hole. But for entrepreneur, the last part of this show is going to be a little entrepreneurial minded. If you were really sick, you would make a feedback form that was money gated. So it's like, cool, I want this thing. I'm willing to pay $100 for this thing. I've paid $100. And once they pay, it goes through the AI workflow it generates the AI generates the feature. And what are you then getting? You're essentially getting custom subscription apps. So right now you can custom apps or subscription apps like, oh, I make this one thing and sell it to a lot of people who pay on a monthly basis or I make this one really expensive thing that I just sell to this one person or one company. Now you're saying like, cool, well, I've, you know, you've requested these features, you just keep paying, but it's custom to you and you really the only person that have access to this. And so soon you might have a couple hundred versions of that same app, but customized. And some people may be paying you $10 a month and some people might be paying you $200 a month.

Right. And the cost to do that to me again is like the marginal cost is near zero. Then that's the, that's the, that's the difference that it opens up a world where you can profit from, from software in a very different way.

AI makes software even more profitable.

Even more profitable, but for very niche. But you could, because it opens up the ability to do custom one to one software in a way where the economics of that are still profitable. Yeah, yeah.

Because look, you and I have both worked in software for a long, long time. We understand the software business pretty well.

We old, very old.

I know feels like I'm old, I.

Feel that way today.

But what you hear in any kind of customer meeting in a software business is oh, I want these things and some other customer wants these other set of things. And you have a whole debate of like, well, what, what did the majority of people want? And if you're not in the majority, we're probably not going to do the thing you want. Right. And now we're basically saying we're going to break that, that. Or it's like, oh, we're going to do build stuff for the majority. Or you could also have this custom app that is your own version of our app with these specific requests that you want that's like long term likely going to be one of the ways software like manifests itself.

Right. We can show two other very quick.

Examples people are probably looking for like, okay, cool, you gave me how to do this, you gave me some of the economics of why I'd want to do this. But like, what should I do? And Maybe show some inspiration.

So this was a really cool one. Again coming back to the ability for anyone to create one to one software. This is someone creating a maths quiz for a seventh grader. It really did. Again that makes you think, right? Like if you are a parent, um, you can actually build apps to help your, you know, child learn whatever they they're trying to learn. So again like someone, a non, a non developer. Again if you're not watching on YouTube, you really should be. But it's the screen is showing a math quiz that someone created for their seventh grader to be able to help them with their homework. I thought that was a really use case. Good, cool use case. The other one that I want to quickly show is, you know there's a ton of like web builders for personal use.

Yes.

Where you can build a lightweight app or a lightweight website. Here again is the ability now to you can create a website on Claude. So we have someone here that's actually going to create a website and publish it on Claude in a very short amount of time. So we have someone here is actually building a website in real time. Right. It's a very simplistic website. But again if you actually think about a personal brand website or these single use case websites, there's businesses and I've looked into them that have a ton of growth for these kind of lightweight website builders that allow you to build something really basic and publish it. I don't see how Claude can't eat up that market because now I can just go in and have on our screen we have someone building a code wizard website and you can see they're just asking it to do all the code and telling Claude what they want and then they can publish it. And just the publish it means that I can start to send it and use it. Right. And if they add in some basic web functionality like forms and these different things, then again they can actually get into the space of these kind of lightweight apps. Lightweight website builders and they can eat up a ton of market share. That is a very profitable space to live.

I'm going to make a bold prediction for you, Karen. You ready to close out the show? In 10 years we will have 10x the websites on the Internet. Because the biggest problem with technical work over the last 20 years is like, oh, it's expensive. And so I spend all this time thinking about an idea, refining my idea, trying to make sure that when I build the thing it's gonna be worth it. And now everything is just gonna get built. It's gonna be like, great, well, I got this idea. I'll just give it a try and see if people like it. And that's gonna have a bunch of good and positive consequences. It's gonna have a bunch of good or terrible and bad consequences. But what's gonna happen is the velocity of just building and making stuff, as we've said before in the show, is just about to grow at another level of exponent.

Yeah, Everyone is the builder. Anyone can bring their ideas to life. It's going to be a game of who has the best ideas and who can get them to market and who can make users happier.

Best ideas. Listen to customers. Really good at distribution. Marketing.

Sales, Exactly.

That's. That's going to be the game. That's Cloud Apps, everyone. Cloud Apps is awesome. We're going to, I'm sure, be showing a lot of other apps that we work on that friends of the show are working on. So this will not be the last time you see Cloud Apps, but we wanted to give you a quick intro today. Big thanks for Kieran for walking through his MVP app and I'm sure he'll be back with a actual live, working version to share in the future. Kieran, let's do it. See you, everybody.

Let's go building some apps.

App time. This data is wrong every freaking time.

Have you heard of HubSpot? HubSpot is a CRM platform where everything is fully integrated. Whoa.

I can see the client's whole history. Calls, support tickets, emails. And here's a task from 3 days.

Ago I totally missed HubSpot.

Grow better.