PortalsOS

The success of LLMs in language tasks was surprising, as language was previously considered different from other AI tasks.

Determining the obscurity of a crossword clue is subjective. Editors consider whether a clue is intentionally featured or simply fits the puzzle. A team of seven editors helps ensure a balanced approach, aiming for accessibility across diverse solvers.

In puzzle editing, there's a need to draw a line with 'Scrabble words'—those that appear only in unabridged dictionaries. This helps ensure that players can find as many words as possible without getting stuck on obscure terms.

In crossword puzzles, if a vertical word is particularly difficult, the horizontal clues should be easier to ensure fairness and solvability. This balance helps solvers find a way in, even with challenging elements.

The Spelling Bee puzzle involves creating words from seven letters arranged in a hexagon. Editors use a database to identify potential words and then curate which words are acceptable, balancing data-driven insights with human judgment.

A good puzzle creates a sense of accomplishment. Whether it's the daily jolt from Wordle or cracking a tough crossword theme, feeling achievement is key.

The inclusion of birds in the spelling bee is debated. Some birds, like the Anhinga, are common in places like Florida but aren't included, which frustrates bird enthusiasts.

There is concern that AI language models (LLMs) are converging and becoming similar, lacking sustainable technical advantages.

The New York Times receives upwards of 150 crossword submissions each week from around the world. Editors sort these into categories: immediate rejections, maybes, and accepted puzzles. The editing process primarily involves changing clues rather than altering the puzzle's words.

Family members often play New York Times games together, turning it into a ritual where they solve puzzles and discuss the words that should or shouldn't be included.