It's not magic, it's text prediction

See what changes when the same model gets few or many clues.

Before the tricks, let's pop open the model's hood. 🐾

  1. Few clues

    The request

    Write an email.

    The AI's answer

    Hello, I hope you are well. I am writing about an important matter. Best regards.

    With no recipient, subject, or goal, it produces a safe, generic continuation.

Example 1 of 2