When writing something, we often make mistakes and want the AI to write it, but AI also makes mistakes. The best option is for the AI to predict the next word; if a human wants, they can keep that or write their own version with a AI companion.
The video is tested with dumb models. The better model you use, the better response it gives. The models I have used in video are under 1 billion parameters.
This project is just to show that AI can make our typing better and make easy typing, and also can make a macOS app. This was a test, not a full product. The GitHub repo is just to test that it can work.
- Because small models are juvenile, they can PREDICT very unnecessary, hallucinated answers. use A better model, like Gemma 4.
Currently this project is immature AND JUST FOR WEEKEND FUN TEST
Log in or sign up for Devpost to join the conversation.