# ML Code Archives

Snapshots of select educational ML projects worth reading line-by-line.

## Contents

### nanoGPT (Karpathy)
Minimal GPT-2 reproduction in PyTorch. ~300 lines.
Great for understanding how GPT works.

### llm.c (Karpathy)
Train GPT-2 in pure C. No PyTorch, no Python overhead.

### minGPT (Karpathy)
The original "minimal GPT" before nanoGPT.

### tinygrad
Sub-1000 line PyTorch alternative.

## Why these?
These projects taught a generation of engineers how transformers
actually work, not just how to use them.

## Source
GitHub source archives of original repositories.