Today's rows: Squishy, Bouncy, Round. Columns: Found at the circus, From a movie, Found at a bakery.

Fill the 3×3 grid with one answer per square that fits both its row and its column; Jev, TypeSafe's AI judge, scores each answer. Free, resets at 00:00 UTC: https://overlaple.com/?ref=lemmy-thread

Paste your results in the comments, and tell us which square you liked best.

[–] [S] 1 point 1 week ago

Thanks, that's a good point. Part of it is already there: when a square locks, the panel under the board shows how strongly Jev thinks your answer fits each category on its own (say 90% yes for one and 30% for the other), and the points are the two multiplied. What it doesn't do yet is say why in words, or make the weak one stand out. I'll look at that.

  • source
  • parent
  • context
  •  

    I'm a solo dev and this is my daily word puzzle, Overlaple. Every square sits where a row category meets a column category (Pirates × Treasure, Space × Transport…), and you type one real thing that fits both.

    Scored on how well your answer fits, not how obscure it is. Upfront, since people ask: the scoring is done by Jev, TypeSafe's AI judge. There's no answer list, so it rates how well your answer fits each category from 0 to 100, and 600 of 900 points wins the day. Every puzzle is tested against the judge before it goes live.

    Free, no signup, no ads, about five minutes. Past days can be replayed as practice. I'd love to hear which scores felt wrong; that's what I'm tuning.

     

    I made this. Each square is where a row category meets a column category (Pirates × Treasure, say), and you type one thing that fits both. It's scored on fit, not on how obscure your answer is: Jev, TypeSafe's AI judge, rates how well it fits each category from 0 to 100, so there's no fixed answer list. One puzzle a day, closes at midnight UTC, and past days replay as practice. No account, no ads. Feedback on unfair scores welcome.

     

    Hi! Overlaple is one grid a day. The rows and columns are categories (Friday's were Doctors / Chefs / Firefighters × Uniform / Tool / Danger), and each square wants one thing that fits both. There's no fixed answer list: Jev, TypeSafe's AI judge, scores each answer from 0 to 100, so a clever partial fit still scores. 600 of 900 wins. Free, no signup, about five minutes. It's new, and I'd really like to hear where the scoring felt unfair.