#machine-learning

Andrej Karpathy's 200-line GPT implementation

Turns out intelligence at its core is simpler than you'd expect. Here's what a 200-line GPT shows you when you slow down and look carefully.