Oct 9, 2025, 4:45 PM

Interested that only 2% fell prey to the dark-Purple mislead. Notepad kept things clear for me. Also interested that a generative model (presumably) with Connections in its training set could so utterly fail to identify the mislead’s category.
#Connections
Puzzle #851
🟪🟪🟪🟪
🟦🟦🟦🟦
🟨🟨🟨🟨
🟩🟩🟩🟩

Skill 98/99
Uniqueness 1 in 72

Screenshot of a segment of the list of common errors, showing that 2% of players played the words limestone, figment, dateline, and plumber—with the LLM-generated category of “Random Nouns?”