Lesson 01
Next-token prediction
It’s guessing the next word. That’s the whole trick.
A language model doesn’t look up answers. It reads what’s there, guesses the next little chunk, and repeats. That’s the entire product.
Pick any one. The unlikely ones are the interesting ones.
Stopped at 12 tokens. Past here, nothing was authored.
Prompt