Context Is All You Have: How LLM Attention Actually Works

You've seen the marketing: "128k context window!" "1 million tokens!" But what does that actually...