I Tried Building My Own AI… Here’s What Actually Happened

A few days ago, I decided to stop just using AI and finally try building one myself. No full roadmap. No perfect plan. Just curiosity and the willingness to figure things out. What followed was a mix of confusion, frustration, small wins, and one big realization: Building AI is way harder—and way more rewarding—than it looks. 💡 Why I Started I’ve been using AI tools for a while, like most developers. But at some point, I kept wondering: How do these models actually connect? What’s happening behind the API calls? Can I build something like this myself? Instead of watching another tutorial, I decided to just start building. ⚙️ The Stack (What I Used) I kept things simple (or at least I tried to): AI Models via API (LLMs) A basic frontend interface Deployment on a cloud platform (like Vercel) Lots of trial and error 😅 Nothing fancy—but enough to build something real. 😵 The Problems I Faced Let’s be honest: things broke. A lot. Some of the errors I ran into: API Error: No endpoints foun