By skyforbes I tried Gemini 3 Pro through mgx, just typed a few sentences describing what I wanted, like a tool to learn French, and within minutes it generated a working web app.
What blew my mind is that it didn't need any external APIs. Pronunciation, word explanations, images, everything was handled inside the model. In a traditional setup, I would have needed at least three or four paid APIs for that.
Even crazier, I could instantly tweak it to learn German, Chinese, or Italian. It felt like having Duolingo, Reverso, and Google Translate fused into one personal app.
It seems like AI is finally moving from assistive coding into actual product creation.
Not gonna lie, Gemini 3 Pro feels different from previous AI models. It makes you want to build another thing, just to see what it can pull off next. BTW, I use Gemini through mgx just to cut corners. It saves me the hassle of deployment (not promoting it). Of course, the real hero here is the Gemini model itself!
Is anyone else experimenting with AI built tools like this? What cool stuff have you built?