Greg Eisenberg joins Sam to break down OpenAI’s new ChatGPT App Store, announced at Dev Day, and why it’s different from the failed custom GPTs. They walk through five specific app ideas — AI tax guy, healthcare concierge, meme maker, AI grandma, and credit score repair — and discuss how a solo developer could build and launch one within 30 days.
Speakers: Sam Parr (host), Greg Eisenberg (guest, app builder / IdeaBrowser)
Introduction: The ChatGPT App Store Moment [00:00:00]
Sam: But it is possible that this becomes a big thing. And if it becomes a big thing, I think this is probably one of the bigger wealth creation moments in AI.
Sam: So we’re doing a special pod because Greg was on the pod where he showed me all these cool tools he was using. Then at the end of the pod he was like, “By the way, did you see that ChatGPT now has a new app store?” And I said, “No, I didn’t. Let’s just record — I want you to show me what’s going on with the new ChatGPT store.”
Sam: And because this is My First Million, can you tell me a bunch of interesting business ideas you see within the realm of the ChatGPT store? A lot of people don’t know this, but when the App Store first came out, a bunch of the biggest businesses came about because they were early. They were first.
What Was Announced at Dev Day [00:01:30]
Greg: That is true. So let me give you a breakdown of what happened, why it’s important, and I can share a bunch of ideas that I think people should steal.
Greg: Let me first talk about what was announced. So the other day Sam Altman talked — it was Dev Day, OpenAI’s big developer conference. He launched the ChatGPT App Store. He gives this example of using Figma in ChatGPT, which is the design tool, and he says “turn this sketch” — something he drew out — “into a workable diagram,” and it was able to do it.
Sam: Wow.
Greg: Yeah, it’s absolutely crazy. So the ability to use apps within ChatGPT basically turns ChatGPT into like a mini internet. You don’t need to leave ChatGPT if you can just connect to apps.
Greg: The other example his team gave was Coursera, the course platform. “Coursera, can you teach me something about machine learning?” It connects to Coursera — you do have to log into it. So if you’re watching this on video, you can see how ChatGPT uses the data. Apps may introduce risk; data is shared with the apps. And then it creates this canvas where you can actually consume content in a more visual way. ChatGPT used to be only text — now it has this canvas.
Greg: There are two ways to find these applications. One is you add it manually, and the other — the way I’m way more excited about — is you just use ChatGPT normally and say, imagine I didn’t add Coursera and I said “I want to learn about machine learning” and it just brings up Coursera as a potential app. That’s the opportunity. If someone listening to this can actually build an app so that it shows up when you ask it a question — Sam Altman just announced ChatGPT has 800 million weekly active users. You can get in front of 800 million people every single week and sell them something.
Sam: Wow. Okay, this is awesome. I just typed into ChatGPT, “How do I turn a drawing into a Figma?” and then I added the Figma app and it told me how to do it. That’s awesome.
Live Demo: Logo Design with Canva [00:05:00]
Greg: Yeah. I’ll just do one live for fun. “Design a new logo for my podcast, The Startup Ideas Podcast.” I forgot to mention who I want — I think Canva. Yeah, Canva was one of them. Canva, the $47 billion design tool, is one of the launch partners. It connects to the app, you log in, and it literally designs a logo within ChatGPT using Canva’s software.
Sam: I mean, there it is.
Greg: There it is. I would give it a 6.5 out of 10. But you know, it’s something I can iterate on — it’s not something you’re going to one-prompt yet, but look, it says “review and select one to continue editing.”
Sam: The font is so weird on some of them. It looks like it’s misspelled, but it’s not. I’ve never experienced that before. Look at the one on the far right, third from the right.
Greg: Yeah, that sometimes happens with AI design — the font’s a bit wonky. You do have to iterate, it’s going to take multiple prompts. But this isn’t bad. And that took like two seconds. I can make this better. That’s awesome.
Why This Is Different from Custom GPTs [00:08:30]
Sam: So what apps or ideas do you think are interesting right now?
Greg: I’ll tell you, but I do want to give the caveat that ChatGPT a couple years ago launched something called custom GPTs. Do you remember those?
Sam: Yeah, it was a flop. I tried making one. It didn’t work.
Greg: Yeah. And you know why it was a flop and why this is different? No one wants to download apps. No one wants to go to a GPT store, set stuff up, download apps. What makes this so much better is it’s contextual. You just type something into ChatGPT and — you might have to mention it — but if you don’t have to mention it and you just say “I want a logo design” and it pulls up Canva or another app, the discovery opportunity is huge. It’s just easier for the customer because they don’t have to set up anything.
Sam: Yeah. I actually invested in one or two companies where things weren’t going amazing and they pivoted into a ChatGPT plugin and they’re killing it. We’ve talked about WordPress plugins and a bunch of different plugin businesses, and I’m on board with those. So I’m eager to see what ideas you have.
Idea #1: AI Tax Guy [00:11:00]
Greg: Cool. So I’ve got 15 ideas I scraped from IdeaBrowser.com, my app. I want you to pick — these are the highlighted ones, my favorites — but I’m happy to go through any of them.
Sam: It looks like the AI Tax Guy.
Greg: So imagine you could prompt “file my quarterly taxes.” It shows a custom UI with deductions, autofilled forms, one-click filing. It connects to your Stripe, Plaid, and Google Drive. And someone is going to build My Tax Guy in a prompt on ChatGPT. It’s a huge opportunity.
Sam: You know what’s interesting? There’s a guy on Twitter who has an AI app he made for uploading your bank statement.
Greg: BankStatementConverter.com.
Sam: BankStatementConverter.com. How much revenue does that make? Doesn’t it make like 50 or 80 grand a month?
Greg: I think it’s $40,000 to $50,000 a month.
Sam: Okay. So it’s a very niche thing, but it’s a thing where if you need it, you really need it. For a bunch of different reasons, I have many K-1s — which frankly I barely know what a K-1 is, but I know that when my wife is getting ready to do our taxes, she’s like, “I need the K-1 for this, the K-1 for that.” We have like 60 or 80 of them.
Sam: Just finding them in my Gmail takes forever. What happens when you invest in real estate or different projects, you get K-1s, and they don’t all come at once — they come from each vendor at a variety of times, in all different formats. The form is the same but it’s sent as an attachment in an email. What would be intriguing is if you could just say “find all my K-1s in my Gmail fast.” K-1 agent — that’s a great idea. Someone should go build it.
Greg: This dude from BankStatementConverter.com — it’s literally one person. Look at this website. It looks like the Berkshire Hathaway website, but for converting a PDF. It’s awesome. You just click a big button: “Convert to PDF.” And I think what you’re going to see happen with this news is apps like this are going to live within ChatGPT. If you ever saw stories about BankStatementConverter and thought “I wish I started that” — if you’re listening to this, you’re early. Now is the time. You should be building that within ChatGPT’s SDK.
Idea #2: AI Healthcare Concierge [00:17:30]
Sam: All right, what’s another one? Keep going.
Greg: AI Healthcare Concierge. Here’s the prompt: “Find me the best dermatologist near me that takes Blue Cross.” Your app connects to ZocDoc, insurance networks, shows an interactive local map and booking options — and then you just make money with affiliate fees. Or it gets acquired by ZocDoc or one of these big companies. If you can own this high-demand query, it’s worth a lot of money.
Greg: How would you make this? Well, I’m lucky — I have a design firm that builds apps like this. I would just hire my own firm. But if I didn’t have a firm and I’m just a one-person operation, I’m using Claude Code or Cursor and building this out.
Sam: How long would that take?
Greg: Realistically two to three weeks to get something working really well. Then you’d need to get approved by OpenAI, which might take some more time — but maybe four weeks total, 30 days.
Sam: That’s crazy. The other thing I would do — and I’ll talk my own book here because I built this for myself — there’s a button on IdeaBrowser that says “build this idea.” I don’t think I’ve ever shown this publicly. But you can actually go and build out the ad creative, the brand package, the landing page, the content calendar, the email funnel system, the email sequence, the lead magnet — all with optimized prompts.
Greg: Have you ever seen that meme of a big meathead sitting at the computer and he’s like, “Build me a business that makes $100 million in revenue in three days. Don’t make any mistakes. Start now.”
Sam: Yeah, yeah. It still needs a human. The mistake a lot of people make is they’ll use something like this and they’ll be like, “Why aren’t customers coming?” Well, it’s you. This is like all your optimized prompts in one place, specific to each LLM or AI tool — but you still need to review it, make sure it works, and optimize from there.
Sam: So what I would do is first version on IdeaBrowser, probably use Claude, v0, Vercel, get some emails up, and try to launch something in 30 days.
Idea #3: AI Meme Maker [00:23:30]
Sam: All right, what else you got? This is fun.
Greg: So memes are getting more and more popular, especially in an AI age. I think whoever owns the “make a meme about XYZ topic” probably makes a few million dollars a year.
Sam: Can you do this now in ChatGPT or Perplexity?
Greg: You could do it on ChatGPT, it’s just not optimized. And that’s the opportunity — that’s why Dev Day happened. ChatGPT realizes they need apps like this to exist so people stay on the platform more. And they’ve realized they can’t do everything.
Idea #4: AI Grandma [00:25:30]
Sam: So what’s AI Grandma?
Greg: This is a fun one. I think a lot of people are feeling overwhelmed right now and they might not want Andrew Huberman advice or even my advice. Sometimes we just need an AI grandma to give us old-school, warm, blunt, human advice. There’s an opportunity to build like a life advisor — you build a brand around an 85-year-old woman who bakes apple pies.
Sam: Have you built any of these yet?
Greg: We’re actively building right now.
Sam: For yourself or clients?
Greg: Both.
Sam: And that’s going to take two to three weeks to get the MVP?
Greg: Yeah, MVP. It depends what you’re building — AI Grandma could take a few days, but something like the K-1 one is more complicated. It can take anywhere from days to, for an MVP, 30 days.
Idea #5: Credit Score Repair [00:28:00]
Sam: Let’s talk about the credit score one, man. Up until recently I had a very low credit score because I didn’t use too much credit. I didn’t realize that at one point, when I had a job years and years ago, I had to take an ambulance ride and I didn’t realize they didn’t pay it. My credit got dinged. I had a credit score I never checked — like 580 — and I was like, what the hell. When I wanted to get a mortgage for a house, I couldn’t. I was searching for “how do I make my credit better? Is there a firm I can hire to just do the work for me?” It was impossible, and the credit repair industry was incredibly scammy. So I’m eager to see how you can build this.
Greg: I had the same problem, which is why I added this. I actually don’t have a great credit score — it says fair, but I’ve started and sold companies. I should have a good credit score. Banks and companies like Experian, they’re just not at the forefront of technology.
Greg: I think there are tons of people asking “how do I fix my credit score?” — you can go check that keyword volume, it’s worth so much money. If you can have something that detects that intent and surfaces your app, I think you need a really good domain. Like RepairMyCreditScore.com — that’s probably taken, but something like that. One of the reasons BankStatementConverter.com worked is because the domain was really good. You might have to spend $10,000 to $20,000 on a domain. But this is one of the more no-brainer examples — it connects to Equifax or Experian, there’s an interactive UI with dispute templates, score simulation sliders, next-step actions, and then you do referral fees or charge for it.
Sam: Does .ai do just as good as .com? From an SEO perspective, from an app perspective?
Greg: SEO-wise — .com is Manhattan. .ai is Williamsburg. Still cool, but Manhattan is Manhattan.
Sam: Well, RepairMyCreditScore.ai is available, my friend.
Greg: Oh, is it?
Sam: Yeah. Why don’t we buy the domain and give it to someone who likes and comments on this YouTube video?
Greg: I can do that.
Sam: Or something like that — just inspire people to actually go build apps.
Greg: That’s a good idea. I’ll buy it and give it to someone.
Sam: Yeah. It’s $50 for year one. I have to use Perplexity to find the discount code.
Greg: And you don’t own any equity — you’re just giving it to them for free?
Sam: I — they can have it. It’s just a $50 gift from MFM to them.
Greg: If they get to a million dollars a year, what do they have to do for you?
Sam: A $100 DoorDash gift card would suffice.
Greg: A 2x return.
Sam: Yeah. A DoorDash gift card. If I can just get about four pounds of unbreaded chicken nuggets from Chick-fil-A, that would have been worth the time.
Wrap-Up [00:36:00]
Greg: Cool. Well, thanks for having me on, and I hope you’re inspired.
Sam: Thank you, Greg. That’s it. That’s the pod.