Startup #1: AI RoastBot

Product Overview

Description: A simple GPT Wrapper that roasts any given image as movie/TV show characters, and personalities.

Initial Thoughts

This is a special project since it was my first startup idea that actually got implemented all the way through and deployed on the internet. It's a silly website that isn't really made to make any money, but I'm glad this is how I started.

Development Process

It took me around 2 weeks to build the whole thing. I hadn't coded in about 2 months, so I was a little rusty when I started, but with the help of the best AI tools out there currently -like Cursor, v0, Claude- it was easy to get right back into it.

During the development, for the first time since AI came out, I felt how efficient AI can make you. I had been using Github Copilot for a while before switching over to Cursor, but in my opinion, it was never more than just a fancy autocomplete. Cursor is something else...

I think dummies on X or Reddit that keep saying shit like,

"AI is just a fancy search engine bro"

should give this thing a spin with Claude 3.5 Sonnet. I really do believe AI will make even more insane improvements within the next few years and change the world forever. Anyway, enough on that for now...

Tools and Technologies Used

I will keep using my simple Nextjs boilerplate until I need something more capable, but here's my stack for this project:

Nextjs
Tailwind with Shadcn
Supabase
Stripe
Amazon S3
Hostinger VPS with Coolify

I'm going to host all my projects/startups on a VPS from now on. I didn't want to pay Vercel $20, and run the risk of waking up to a $96,000 bill.

Lessons Learned

The main thing this project taught me is I need to launch faster and faster. It's so much easier to launch a decent project nowadays; it's a joke.

Something more technical; I used S3 buckets for the first time for image storing and although setting it up is kind of a mess, it's dirt cheap and honestly not that diffucult. I will probably use S3 in the future sometime soon.

I also had to do some prompt engineering when it came to making sure the AI actually does roast the image given. I'm sure I could make it even better, but it works about 90% of the time, which is good enough for now.

Next Steps

Since this was more of a 'let me get my toes wet' kind of project, I don't plan on marketing it much or improving it any further anytime soon.

I just want to build more projects and launch quickly.

I already have some ideas in mind -actual useful ideas- so I will pick one and start coding the next thing as soon as possible.

Check out AI RoastBot though and let me know what you think...