Heyo! I thought I’d share the story of how I created FlickPal.io and some tips I picked up while using Replit.
I’ve worked in technology for over 10 years, mostly with Microsoft products. I’ll admit, aside from dabbling with Microsoft 365 Copilot, I hadn’t explored much of the AI world. Over the Christmas break, I decided to change that and dive into AI models and software to see what was out there.
While listening to the MFM Podcast, Amjad Masad, spoke about the platform so passionately that I had to try it out. Right away, I was impressed. Not only by the AI capabilities, but by how easy it was to get started with software development.
Throughout my professional career, I’ve seen so much time wasted on “Sprint 0,” setting up development environments and deployment pipelines. Replit just skips all of that, which was a game-changer for me.
Setting myself a challenge
I gave myself a goal: build an AI-powered movie recommendation website in under 24 hours and I wanted to do it using ONLY Replit’s Agent and Assistant, without making any code changes myself.
The Idea
The first thing I did was figure out what makes a good movie recommendation. I realized it’s not just about finding a great movie but finding one you can actually watch. There’s no point recommending something that’s only available on Hulu if you only have Netflix.
From there, I mapped out what the app would need to ask users for in order to generate meaningful recommendations. Here’s the basic structure I came up with:
- Streaming Service: A list of predefined platforms for users to pick from.
- Genres: I created a table with 50 genres and had the app randomly load 15 of them each time. Users could also add their own genres if needed.
- Description: A free-text field where users could provide details about what they’re looking for.
- Content Type: Whether the user wants to watch a movie or a TV show.
- Watch Time: This one came after some testing. Let’s face it, we don’t always have unlimited time, so being able to filter based on how long you have is super useful.
- Featured Actors/Actresses: Similar to genres, I created a table of 100 actors and actresses (all generated with Replit) and had the app load 15 randomly. Users could also add their own favorites.
Choosing the Tech Stack
Replit’s Agent and I discussed, and settled on this tech stack:
- Front-End:
- React with TypeScript
- Material-UI for components
- Back-End:
- Express
- Drizzle ORM
- Database: PostgreSQL
- AI: OpenAI GPT-4
What I Learned
Here are a few key things I discovered during the process:
-
The Agent is great for the big stuff. It’s awesome for kicking off projects and building major features quickly.
-
Stick with the Assistant for minor changes. The Assistant is helpful for debugging or tweaking code, but don’t blindly trust its suggestions. Sometimes it introduces errors that can set you back. I started reviewing its suggestions carefully and, when they didn’t look right, prompted it to adjust. It can be a little soul destroying having your app mostly working, and then making a change with assistant that breaks it all. So, reviewing the files and what it suggests to change is really important. I find this is also a really great way to learn the codebase, so you can self manage later if you need to.
-
Chatto the Assistant. Seriously, treat it like a teammate. I constantly asked it about the codebase or server behavior. At one point, I suspected duplicate API requests and asked the Assistant to look into it. It confirmed the issue and suggested a solution, which was super helpful.
Bringing everything together was such a rewarding experience. It’s amazing what you can accomplish when you dive in and let the tools do some of the heavy lifting.
FlickPal.io is live now and free to use, so give it a try if you’re hunting for your next movie or show to watch. Feedback is also most welcomed. 