About Contexto

Contexto is a daily word game. One secret word each day, unlimited guesses, and the only feedback is how close your guess is in meaning.

How the ranking works

Every word in a 70,000-word vocabulary is scored against the day’s secret word using a machine-learning model trained on how words are used together in ordinary text. The words are then sorted by that score, and the position you see is the rank. Rank 1 is the answer itself.

This is why meaningmatters and spelling does not: for the word “ocean”, the words “sea” and “beach” rank in the double digits, while “occasion” ranks in the tens of thousands.

How a puzzle is actually made

Every language is built separately, from scratch, and none of it is translated. The pipeline starts from word vectors — numerical representations of how words are used across a very large amount of ordinary writing — and from those it assembles that language’s own vocabulary. Nothing crosses between languages, because the words closest in meaning to ocean are not the translations of the words closest to océano.

From that vocabulary a smaller set of answers is chosen by hand, and then every word is scored and ranked against every answer in advance. By the time a puzzle goes live the full ordering already exists, which is why a guess comes back instantly and why the rank you are shown is the same rank everyone else is shown.

Puzzles are prepared months ahead on a fixed schedule, and a separate pool of practice words is held back from that schedule entirely — so unlimited mode can never hand you a word that later turns up as a daily.

Fairness

Answers are drawn from a hand-reviewed list of common, concrete nouns — roughly nine in ten are things you can picture. That is a playability decision rather than a squeamish one: when the answer is abstract, the top of the ranked list fills with near-synonyms that all feel equally close and the last few guesses become a shrug.

Proper nouns, misspellings, slurs and distressing subject matter are excluded from the answer list by a combination of automated filters and review. Some of those words remain guessable, so that a reasonable guess still gets a real rank.

Corrections

Puzzles are generated, so mistakes are possible — a word that is ambiguous in a way the model did not capture, or a rank that does not match how people actually use a word. When one is reported and confirmed, the word is pulled from future rotation. Published answer pages are corrected in place rather than quietly deleted. The full policy is on the editorial policy page.

How the site is paid for

Advertising, and nothing else. There is no subscription, no paid hint tier and nothing to buy — the game and every answer, hint and archive page are free and always have been. There is no account either, so playing does not require handing over an email address. What is measured, and how to opt out, is set out in the privacy policy.

Who runs it

One person, independently. It is not a studio, a network site or a content farm, and nothing here is generated on demand — every puzzle is prepared and reviewed before it is published. Email reaches that person directly rather than a support queue.

Independence

Contexto is an independent project and is not affiliated with, endorsed by, or connected to contexto.me or its creators.

Contact

Corrections and feedback: hello@contexto.us.com. See the contact page.