How I Build
In the last 20 days, I’ve created around 8+ apps, each in a matter of hours. For example, OOPSDEAD, closet.gg, OpenMagic Pro, openfantasy.app, sourceOS, and policyOS. I even created two concepts for generative AI NSFW websites: NSFW1 and NSFW2.
Here’s roughly how I went about it. I’m not sure if I’ll be updating this post in the future, so it’s possible the websites above won’t age well.
1. Idea Generation
I start by brainstorming ideas using a voice recording app on my phone.
2. Project Creation
I upload the voice note to bld.ai, which creates a structured project for me. If needed, I refine the output using ChatGPT 4.5.
3. Prototype Development
I paste the project brief into lovable.dev, which generates a front-end prototype
- Frontend: Tailwind CSS
- Backend: Supabase
- Framework: Next.js
- Deployment: Vercel
4. AI Integration
For AI functionality, I deploy models using Dokku on a Hetzner cloud server.
- LLM: OpenAI GPT-4o
- Image: Stable Diffusion via PromptEngineering
- Voice: ElevenLabs
5. Project Scaling
Sooner or later, AI Agents aren’t become tedius and frustrating to use. I want a smart, hard working, and detailed oriented team of hoomans. And I want a framework to scope, budget, and deliver the work. For that, I use bld.ai.
I’m having fun. I hope you will too.