Building with AI + Replit: how to rapidly build and deploy AI-powered apps
Matt Palmer from Replit shares his process for turning ideas into applications—fast.
Published 2025-03-21
.jpg)
If you're in the AI space, you've probably come across Matt Palmer on X (Twitter). He’s constantly sharing innovative builds, clever workflows, and insights on AI development.
As the Head of Developer Relations at Replit, Matt spent an hour with us demonstrating how to build powerful applications with Replit—all from your browser.
In the workshop, Matt took us through his process for turning an idea into a working app using Replit. If you're looking for a way to build AI-driven apps quickly and without coding expertise, his session is a must-watch.
Here’s a taster of what Matt covered with us:
Going from an idea to a live app
First, Matt explained how Replit differs from other tools:
“Replit is a little bit different because it runs entirely in your browser, and that means there’s no installation. It’s all in one tab. And we’ve added on AI tools that can help you every step of the way, like Agent and Assistant.”
For his demo, Matt started with a simple concept—building an interactive map of San Francisco’s privately owned public open spaces—and showed us how to break down the project into manageable steps.
Before getting the AI to write a single line of code, Matt emphasized the importance of breaking down the different technical requirements for an app. He outlined what he needed:
- The core features needed (interactive map, search, filters, customizable markers)
- The tech stack (React for frontend, Leaflet for mapping)
- Data sources (public APIs for location data and artwork in SF)
Matt showed how giving AI structured context—like relevant APIs and documentation—led to a much smoother development process. As he explained:
“What I’m doing is breaking down all the functionality that we’ll need and trying to visualize what we’re gonna build... If you know the technologies that you’re gonna use and you know that they work with AI or you can give them context, you’re gonna get a better result.”
Using Replit’s AI-powered tools: agent & assistant
Matt introduced us to the newly rolled out Replit Agent, an AI developer that can set up full-stack projects from a simple prompt.
Unlike other AI coding tools that generate isolated scripts, Agent creates entire project structures with frontends, backends, and deployment-ready files. Matt explained that “Replit is a workspace and Agent is our tool that builds on top of the workspace.”
Once the initial project was generated, he used Assistant to refine and modify specific parts of the code. Agent is best for major project set up and large feature additions, while Assistant is better suited for debugging and making small tweaks.
Matt described Assistant as a real-time coding partner that speeds up iteration, allowing developers to adjust their applications efficiently.
He clarified when to use Agent vs Assistant, saying:
Building with AI: debugging & iterating
Matt made it clear that building with AI is not a “one-and-done” process—it requires problem-solving and iteration. As he worked through the app, he encountered several issues:
- Unexpected UI issues: The search bar overlapped with the title.
- Missing features: The map lacked visual distinction between spaces and public art.
- Data integration bugs: Some location markers weren’t displaying correctly.
Instead of manually digging through code, Matt explained how Assistant could fix specific issues when given targeted context: “Assistant knows about all the files ... but when I mention a file specifically, it’s going to know to hone in on that.”
Once the app was “85 or 90% of the way there”, Matt mentioned some potential enhancements, then deployed the app, explaining that Replit will take care of all of the build, deploy stuff so you don't have to worry about it.”
Keeping costs under control when using AI
One question toward the end of Matt’s session was about databases, specifically “how to think about costs and how do you prevent building something that gives you a crazy bill at the end of the month?”
Matt went through some best practices for managing costs and database usage, especially for AI-powered apps that make frequent API calls. He recommended checking the documentation for your chosen database provider (e.g. Neon) and making sure you understand “how they charge and what they charge for.”
He also demonstrated how, within Replit, you can keep an eye on costs.
“You can see exactly what you're getting billed for. So you can go to storage, you can see which apps are using a lot of storage. You could go to these different things and understand exactly the cost.”
Want to create your own AI full-stack app, without coding?
Matt’s workshop showed how straightforward it is to create a full-stack app using Replit, iterating on and fixing bugs with the help of AI. Want to check it out for yourself? Watch the full recording of Matt’s session (available to Pro Ben’s Bites members).
Liked this post? You may be interested in these: