How answers are scored
Every answer is worth 0 to 100 points. Common answers score near zero. Rare ones score near 100. Each prompt also has one chosen summit answer, worth exactly 100, so reaching 700 means finding every summit.
Rarity is worked out in advance, before you play. It blends how often a word shows up in everyday writing with an estimate of how likely people are to say it first for that prompt. It never looks at what other players typed, and your answers are not sent anywhere.
Modes
- Daily: the same seven prompts for everyone, once per date. Your result is kept in your own browser.
- Archive: every past daily, each playable once.
- Unlimited: five random prompts a round, as many rounds as you like.
Where the answers come from
Lists of valid answers are built from Wikidata (CC0), checked and trimmed with AI assistance and corrected as players report mistakes, with word frequencies from the open-source wordfreq project (data under CC BY-SA 4.0). The art, wording and code are original to this game.
Something wrong?
If a real answer was rejected, a silly one was accepted, or a score looks off, please tell us on the contact page. Bad data is the thing we most want to fix.
How Thin Air is made
Thin Air is a small independent project. It was built with the help of AI tools (Anthropic’s Claude), which wrote much of the code, the first drafts of the answer lists and ratings, and the first drafts of the guides on this site. The owner decides what the game is, tests it, and edits and corrects it.
Early players shaped it. After the first friends tried it and found some questions too specific, we moved the narrow ones (such as exact letter counts) out of the daily game and wrote many broader, everyday ones. Players are still the best source of fixes: if a real answer is rejected, a score looks off or a prompt feels too narrow, tell us.
Scores never use what other players typed. They come from fixed ratings, explained on the scoring page and in rarity without a crowd.