AI (best version)

I built an AI from scratch using numpy. I had no idea how LLMs really worked at the time, so I guessed, creating this single-layered AI that runs on a simple sigmoid function. It does not predict tokens, instead, it takes 5 characters as input and outputs 5 characters. It works surprisingly well for its size and it was able to understand more than I taught it, which proves that it wasn't too overfit. After this, I learned how LLMs really worked and went on to create some smaller text prediction models of my own.