Micrograd

Lecture: https://www.youtube.com/watch?v=VMj-3S1tku0 by Andrej Karpathy
Notebook: https://deepnote.com/workspace/kunal-bhalla-372e292e-4a71-42dc-b95a-fb0bf5e63214/project/Untitled-project-971b8394-46af-46aa-8e03-3ea5ef9279e1/%2Flecture.ipynb
Neural networks are a mathematical equation
- take the weights as an input
- no one actively derives the derivative in a nn
Backpropagation is general and can deal with any math
Heh, I did an even more minimal version of this in Rust some years ago

*