How it works

Every day at midnight Eastern time, a new set of six problems of the day goes up, drawn at random from twelve categories. Everyone gets the same problems, and your first attempt at each counts for the daily leaderboard. The clock starts when you open a problem.

Every puzzle is generated from a seed and checked by a solver before anyone sees it: logic grids and Sudoku have exactly one solution, sequences are rejected if a simpler rule fits, and word puzzles are checked against a dictionary.

Practice is unlimited and unranked. Your rating per category uses Glicko-2, treating each puzzle as an opponent; adaptive practice picks the difficulty tier that matches your rating.

Each problem has a par time. Solving it at par earns the problem's base score; solving faster earns more, with no ceiling (four times faster than par doubles it), and slower earns less, down to half. Hints cost 25% of your score each. Memory drills are never ranked, since the answer has to be shown to you.

What we don't claim

The evidence that puzzle training improves general intelligence or prevents cognitive decline is weak. Hardset is for the challenge and for getting better at these specific skills.