Should OpenAI have celebrated humans whose puzzle solutions were one variable off?

seanmcdonaldxyz · x · 2026-10-10

• Commenting on OpenAI's puzzle-solving event, the author proposes a compromise: OpenAI could have cooperated with the humans whose solutions came closest and released them together.
• Core point: if your published solution was just one variable off the model's final answer, you should have been invited to the party.

Original post →