# micrograd

A scalar-valued autograd engine built from scratch, following Karpathy's lecture. The single best way to make backprop stop feeling like magic.

## Status

[ ] Not started.

## Prerequisites

[`prereqs/calculus/`](../../prereqs/calculus/) — chain rule.

## Resources

- Karpathy — *The spelled-out intro to neural networks and backpropagation: building micrograd*.