01 / The Model
Next-token prediction
The only thing a model does. It guesses the next small piece of text, adds it, and guesses again. Every impressive behaviour you have seen is built on top of that one trick, repeated.
How you’ll hear it
“How does it actually come up with an answer?”
“One small piece at a time. It guesses the next bit, adds it, and guesses again.”
Avoid
Describing the model as looking things up or thinking like a person. It generates, one piece at a time.
The longer version
Everything a model produces comes from repeating one step: predict a likely next token given everything so far. Summaries, analysis and code all emerge from that loop. It is also why confident mistakes happen: a fluent next piece is not the same as a correct one.