Synchronous GPT

With so many AI models locked behind separate subscriptions, I figured there had to be a better way to access all of them without paying for five different premium plans. Around the same time, "AI vibe coding" started blowing up online, so I decided to experiment with it and see how far I could push it. I downloaded Cursor IDE and started building a SaaS platform that would let users pay for one subscription and use multiple AI models in a single, unified dashboard.

In about a month, I had a fully functional prototype. It worked well enough on my machine, but the cracks showed pretty quickly: certain features failed often, authentication broke for other users, and some pages simply didn't load properly on different devices. It became obvious that the issues came from relying too heavily on AI-generated code without fully understanding the underlying systems.

So I decided to back up and do it the right way. I enrolled in a full-stack web development course, rebuilt the architecture from scratch, and gained a much deeper understanding of modern web stacks. I'm now actively working on a significantly more stable and scalable version of the platform — one that I actually trust to handle real users and real traffic.

The current build uses:

• Next.js & React.js for the frontend

• JavaScript across the entire stack

• Supabase for authentication and database storage

• Stripe for subscription payments

• Resend for email notifications

• Vercel for deployment

I'm still actively developing the platform and expect to release the improved version soon. As a project, this one has pushed me outside my comfort zone, forced me to actually learn full-stack engineering principles, and taught me the value of understanding the code you ship rather than blindly trusting AI to write it for you.