Lowering the barrier to first charge: How agents are growing the economy
Charting the future of payments
Runtime
Complete form to watch full video
The distance between an idea and a first charge is shrinking fast, and that has implications for the entire economy. Experienced engineers are shipping in hours what used to take weeks. First-time builders with no coding background are turning ideas into businesses. Hear from both, and see what's possible with Replit—from natural language prompt, to working storefront, to real transaction, in minutes.
Speakers
Jacob Eisner, Senior Field Engineer, Replit
JACOB EISNER: Now, I know what you’re thinking. What the hell is some 19-year-old going to tell me about how to implement AI in their business? Well, don’t worry. I’m actually 24. So in all seriousness, I’m Jacob. I’m the founding field engineer hire Replit. I’ve worked with enterprises, individuals, I’ve worked across the whole stack. I’ve worked with over 200 clients, and I’ve helped them implement AI in their businesses. Because of all this experience, what I’ve seen is I’ve seen this elephant from every angle. I’m one of the top users of Replit and Replit. And because of this, I get a lot of the same kind of questions from people. And what I get most often is about who is AI going to disrupt the most. Now, we’ve all thought about this. Some of us think it’s software developers. Others think it’s designers. Maybe SDRs are being brought up, but I actually don’t think it’s any of those.
I actually think it’s venture capital. And the reason for that is in the past three years, the cost for taking a business idea from zero to one has dropped by around a thousand fold. And the reason for that is because people are building their ideas using AI without any developer help, and they’re monetizing them using a Stripe integration within their applications. And what this means is they don’t need to go for venture capital as early. They can get from zero to one without a seed round. So this is a super exciting time to be starting a business in AI. And what we’re going to do today is we’re going to talk a little bit about how you can generate this ROI really quickly using Replit and Stripe.
So why don’t we hop into it? OK. Here I am on Replit. You can see I got a reminder for this session here. So all you have to do is, on Replit, you can generate code, you can generate applications, you can generate software purely through natural language. So what I’m going to do is I’m going to say, “Please make me an eshop for selling pet rocks. Please use Stripe.” OK, cool. And now, as we see here, Replit Agent will start working. Now, what it does is it’s going to take your prompt and turn it into code. Now, when you develop software traditionally, you have to set up your entire development environment. You have to configure everything yourself. You have to know a few things. It’s good to have a computer science background. On Replit and other vibe coding platforms, what we do is we spin up the development environment for you.
We call it a “batteries included” development environment. Here, what you’ll see is our agent will start calling tools, reading docs. It has the ability to search the web to research information about the Stripe integration. And what it’s going to do is it’s going to prompt me to put in my Stripe API key and I can get started. Now this is super simple. It’s going to come up with a small prompt here that says, “Enter your API key.” It’s going to link me to the Stripe documentation. I just have to grab that API key and then I get to enter it into Replit, and then I’m pretty much done. I see people building full applications, pretty highly detailed, in about a day. So most of these happen in an afternoon. It’s super exciting to be part of this and seeing this part of Replit. So let’s go into Replit a little bit more here.
You can see here it’s calling the tools. This is the agent chat on the left. I can send initial prompts. Sometimes prompts are one sentence and sometimes they’re an entire PRD document. Replit can handle both and I can send follow-up prompts here. If we actually pop, press this button here, I can look at the file tree. And this is all the code that the agent is generating off here. So you don’t have to think about that at all. In fact, 90% of the Replit users don’t even pop this file tree ever. You can purely talk to AI.
To be batteries included, you need a few other things with you. You need a database. Agent can provision a database, can connect your database, and then you can do things like store payments, store users, et cetera. Cool. The other thing you need is a secrets manager to securely carry your secrets. We have this here, and this will manage all of our secrets. You’ll see agent is talking about launching a subagent to do the work in parallel. Replit has the ability to do work in parallel. So what it’s going to do now is it’s going to launch a subagent to build out the front-end UI while the main agent works on the back end. OK? Cool. And then the last thing that you need to be batteries included is the ability to publish your application to the outside world. So you’ve built an ecommerce website. You need to deploy it.
And traditionally, that was very hard. But on these vibe coding platforms, what you can do is you can click “Publish” and then share the app to the world. So you’ve gone from zero to one in Replit. I’ve actually heard stories of people raising $5 million rounds purely based off of vibe coded applications now. It’s a very exciting time. Cool. So while this works here, it’s going to wire up the Stripe API. All you have to do is, you can hop into Stripe, get a developer account, set up a business, and then you’ll get these API keys. Now, my advice for when you’re working with Stripe is don’t show your secret key to around 200 people.
So we’ll wait for that to finish, and all we’ll need to do is paste this value in. It literally is a two-minute process. Cool. Well, while it’s cooking, let’s move on to one that I premade here. So here’s a premade app in Replit. This is again, the pet rock application. You can see I have these various pet rocks. This took me around an hour to make. Cool. Now, if I pop this out into a new window here, what I can do is I can see, I can click “Add to manifest,” and then go to my cart. And I have these items in my cart and I can proceed to the purchase, which will take me directly to Stripe. I did nothing more than paste in my Stripe key and automatically I have a functioning business now.
I can make my first $394. Cool. There are a few other things that you’d want to do with AI, right? Perhaps you’re starting out your business. Perhaps you’re building something internally to your company. If I click this drop-down, you can see that it’s also created some slides for a pitch deck to VCs. This is one prompt, and it will use the context from the actual application to build your slide deck. So you can not only have a working payments platform, have revenue, but you can also do your pitches. Our CEO talks about having this all in one platform where individuals can build their entire business. And with Stripe and Replit, you can do that. OK. You’ll notice that this design is quite nice and that’s because I have great taste, but it’s also because I can do rapid ideation on design in Canvas. What I can do is I can select here, I can say, “Show me a few more variants of this design.”
And what it will do is it’ll spin up a few more rapid prototypes on this Canvas. You’ll see that this design actually matches this one here. And that’s because what I can do is I can select one of these little previews and then I can click “Build,” and then it will build it into my application. OK. Cool. So here we can see the Stripe API key. All I have to do to input this into my application is click to copy the secret key, paste it in here, and click “Continue.” It looks like it prompted me twice. And now when you paste it in here, the AI has no way to view that secret key, so it’s completely secure. Cool.
OK. So we have a few variants going on the Canvas here. The other thing I can do is I can click plus. So what’s just starting to happen is people are starting to orchestrate multiple agents at the same time. What you can do is you should be thinking about how can I orchestrate? How can I think one layer above just a single agent? If I click this plus, what this lets me do is spin up a second agent. So I can say, “Please add an ‘about me,’ ‘about us’ section to the site.” And now I have two agents working in parallel. You can see this one’s working while that one’s working. The way that’s possible is what we do is we fork the code, we work on the code, and then we merge it back in all automatically, so you don’t have to worry about that.
OK. You can actually have up to 64 agents running in parallel. So you should really be thinking large scale agent orchestration now. OK, cool. I’ve shown you how to build something very basic in Replit, but the next question that should be on your mind is, how do I build something great?
So if we hop back into our presentation here, we’ll walk you through that. How do you make great things using AI? Maybe someone has demoed something for you using AI and you thought, “Man, this is not cool. I don’t care.” The ability to create things is really easy using AI, but the ability to create bad things is also really easy and everyone is creating bad things. So you have to create some good things. So I think I have a three-step framework for this that I usually tell people I work with. The first step is to be audacious. I go to all these hackathons with all these different people and they always build a dashboard. Dashboards are so 2025. Don’t build dashboards, build something cooler. AI is extremely powerful. It’s extremely competent. So ask it for more difficult things, right? Ask for a mobile app. Ask for a desktop app.
Ask for something that lets me train AI within the application. Ask for something that can connect to Twilio and can do real-time streaming of audio. Don’t limit yourself to just making a simple web UI.
Cool. And then I’d say also, you should be resilient. I see a lot of people that don’t believe in AI. And I think the reason for that is because they put in one prompt and it doesn’t work out for them, and they never try again. And then they say, “AI is bad.” And for them, it’s a bit of a self-fulfilling prophecy. If you don’t try again, yeah, AI is bad. But when AI is wrong, you must prompt again. You must constantly think about how you can change your approach. It’s super easy to learn using AI. You may have noticed that we actually have a plan mode. Most AI tools do. What this lets you do is ask any question about the code base. You can ask how code works. You can tell it to teach you things. You should always be learning as much as possible.
And therefore, when it gets something wrong, you know how to fix it. So don’t give up on the first sign of resistance. And I see people doing that a lot. I urge you not to. Cool. And then lastly, be a visionary. So it was Edison that said, “Genius is 1% inspiration and 99% perspiration.” Well, AI can do 98% of that, so get inspired.
I see, you should be thinking about, what’s the ideal end state for my application? Because getting to the end state is almost free. So what is the perfect ideal state for my app? I see a few different types of personas that vibe code. The first person I see is they get access to like an AI coding tool, and they realize they can make anything, so they do make everything. And whatever software they build, it has every single feature. They basically build ERP, right? Submodules, but there’s no vision behind it. So I urge you to be a visionary about what you’re building and not just build every single feature. The other persona I see is more of the designer. They’ll focus in on one part of their application and make sure it’s exactly perfect, pixel perfect. But they’re not seeing the forest of the trees. They need to see the broad picture, work on the important things.
The persona that I see succeeding the most with AI is the product manager and like the CEO founder. The people who can get all of this information about the world and they’re very in touch, but they can condense it into a really good vision for a great product.
So the other things people ask me: how do I write secure code using AI? So let’s talk about security. Security’s a little bit interesting. AI is extremely good at writing very insecure code very quickly, and I can write a lot of it. But the great thing about AI is it also has the ability to remediate a lot of insecure code, and it can do it better than a lot of humans can already. So the way I see it is, there’s multiple ways to implement security, and it’s always a trade-off between engineering effort and how secure you want to be. There’s always multiple approaches. With AI, you can always choose the most secure approach, right? Because now the cost of the engineering effort has gone down to zero. This is something that we actually do in Replit. We call it the “pits of success.” So when a user doesn’t define what security approach to implement for whatever feature they’re implementing, what we’ll automatically do is we’ll choose the most secure implementation, right? Whatever it may be. And we’ll spend a few extra tokens to get there. Therefore, we guide the user into a pit of success where they’re secure and they can fall into this pit of security.
Generally, when you’re prompting, if you’re about to go live, there’s nothing wrong with prompting. Please do an audit for how secure this application is. Do something very in depth. You can go through and reaudit everything every time you make a change because it’s almost free to do so. You should have this abundance mindset when it comes to security. I see people prompting, “Make me an architecture diagram for security.” I also see people prompting stuff like, “Do a pen test of my application. Write some code that can actually interact and potentially penetrate my app to prove that it’s actually secure.” So think in terms of abundance. And then the question is, how do you write scalable code using AI? This is the next thing people ask me.
On Replit platform, the most I’ve seen, the most requests per minute I’ve seen is around 50,000. So that’s 50,000 requests per minute hammering one app at once. And it was handling it just fine. Usually scalability comes a lot later than you think it does. But again, you should take the same approach to security. Do a scalability audit. How scalable is my software? You don’t have to worry about these things when you can simply ask the agent to, “Provide me a scalability report,” or “What will break under load for this application?” So ask the AI, don’t be afraid, be resilient, and don’t be afraid of anything going wrong during your applications because you can always continue prompting. Cool. And let me finish you guys off with a story about a successful business owner using Stripe and Replit.
I was on a call with a client, Eman. He has no technical background. He has a background in finance. And I was on a call debugging his app with him, and he was telling me a bit about his business. And what he did was he saw all these AI transcription, translation services coming online and he built his business, whisperai.com. He built it completely on Replit, purely prompting through agent. And he’s slowly grown this business by himself and now he generates over a million dollars of revenue a year. His sole job is vibe coding. And he was only able to do that using vibe coding and using Stripe. So I think it’s a really exciting time, that you can make a big difference in the world today.
So there are two things you guys should do. That is, come up with an idea and get it off the ground using Replit. Second of all, ask the agent to integrate Stripe into your apps and see your first dollar. Happenstance, Replit is free this weekend. So please start prompting. Please start asking your agent to implement Stripe. Thanks for having me guys.