This was so fun! What a delight, thank you.
After the first round I wanted to keep playing, so habitually I just tapped whichever button that most closely resembled ‘play again’, and it took me a couple questions to understand that it was a different mode. Maybe putting “type the OLDEST answer”, or “the deepest cut” above the prompt box would help.
Also the model’s feedback was surprisingly astute! I expected “incorrect”, but got “Help was a top-selling Beatles album, but it sold under 20 million copies.” Love learning during games.
Plenty of good answers were rejected, especially where typos were concerned. On the coffee shop question, CK's coffee shop wasn't found and Waffle House was rejected, but Paul's was a top answer? Paul's also serves breakfast food. Seems arbitrary. Memphis Colosseum was rejected probably because the proper spelling is Coliseum. But how could it not find it from "Memphis Colosseum"? It's easily found (and the spelling corrected) in Google Search.
Hey, thanks a lot for the feedback! Most answers are first checked against a fixed list, to avoid making too many LLM calls. Unfortunately, for latency and playability reasons, we can't do much deeper online searches: every call would cost a lot and take too long. So about 1 to 1.5% of answers end up judged a bit unfairly, but we're actively working on improving the model.
Thanks again for your feedback!
Lots of fun, but one really annoying bug that makes it hard to play: anytime I have an answer that isn’t accepted, and then go to type in a new answer, the keyboard completely covers the text box which makes it impossible to delete or edit the previous answer! Safari on IOS 26.6.2
Thanks a lot for the feedback chrome tells the page exactly where the keyboard is and firefox doesn't... and in that case my layout keeps the question above the visible area. I'll try to fix it
Thank again!
I love that I could justify my answer to it - spoiler below
I chose CompuServe for oldest file sharing service, and it told me that it was an online service not a cloud file hosting service. I then answered "CompuServe had file sharing" and I received 100pts
I'm really glad you liked it! That's one of the big advantages of using LLMs instead of relying on a fixed list of answers. Thanks so much for playing! :)
The agent is too helpful, I got away with stuff like "The one with the southmost shinkansen" for rarest Japanese city, which it substituted for the answer to that (which I didn't know)
Rarest is terrible. I absolutely know that baked ube is a street food, and way less likely to be guessed than Udon, and yet I'm penalized for nobody else guessing it yet?
I didn't even play latest and oldest, because that's just a guessing game.
Thanks you for the feedback, the concept doesn't come from Krillion or from me, there was already a game called Pointless on TV way before that, and Sky is straight up Family Feud, Past and Future are our own concepts
And the graphics, even if the code and the implementation were obviously written with AI like 100% of people coding a big project today, they're human made
But the Abyss mode looks almost identical to krillion - from the very concept of trying to guess rare answers, to the nautical diving theme, down to the almost-pixel-art style..?
It's incredible to suggest that this wasn't based on either Krillion, or that both games were based on something else.
(To be clear, I'm not accusing you of explicitly lifting the concept - but that's the problem with using AI to generate too much, it'll obviously re-use what it's seen before.)
edit: actually, you say "A large part of the code was written with AI tools, but the artistic direction the concept and every choice design are human-made" - it's just wildly improbable for the two to look and feel so similar.
This was so fun! What a delight, thank you. After the first round I wanted to keep playing, so habitually I just tapped whichever button that most closely resembled ‘play again’, and it took me a couple questions to understand that it was a different mode. Maybe putting “type the OLDEST answer”, or “the deepest cut” above the prompt box would help.
Also the model’s feedback was surprisingly astute! I expected “incorrect”, but got “Help was a top-selling Beatles album, but it sold under 20 million copies.” Love learning during games.
Great work, the lot of you!
Plenty of good answers were rejected, especially where typos were concerned. On the coffee shop question, CK's coffee shop wasn't found and Waffle House was rejected, but Paul's was a top answer? Paul's also serves breakfast food. Seems arbitrary. Memphis Colosseum was rejected probably because the proper spelling is Coliseum. But how could it not find it from "Memphis Colosseum"? It's easily found (and the spelling corrected) in Google Search.
Hey, thanks a lot for the feedback! Most answers are first checked against a fixed list, to avoid making too many LLM calls. Unfortunately, for latency and playability reasons, we can't do much deeper online searches: every call would cost a lot and take too long. So about 1 to 1.5% of answers end up judged a bit unfairly, but we're actively working on improving the model. Thanks again for your feedback!
queue jev...
Lots of fun, but one really annoying bug that makes it hard to play: anytime I have an answer that isn’t accepted, and then go to type in a new answer, the keyboard completely covers the text box which makes it impossible to delete or edit the previous answer! Safari on IOS 26.6.2
Thank you so much for reporting this bug! It should be fixed now, let me know if it still happens. Thanks again for playing! :)
Firefox on Android, sometimes the question is off screen after answering a question
The sloppiness shows, the performance is awful (Firefox on Pixel 8)
Thanks a lot for the feedback chrome tells the page exactly where the keyboard is and firefox doesn't... and in that case my layout keeps the question above the visible area. I'll try to fix it Thank again!
Are there no accounts? Can I not continue playing on a different PC?
I love that I could justify my answer to it - spoiler below
I chose CompuServe for oldest file sharing service, and it told me that it was an online service not a cloud file hosting service. I then answered "CompuServe had file sharing" and I received 100pts
I'm really glad you liked it! That's one of the big advantages of using LLMs instead of relying on a fixed list of answers. Thanks so much for playing! :)
It was fun!
The agent is too helpful, I got away with stuff like "The one with the southmost shinkansen" for rarest Japanese city, which it substituted for the answer to that (which I didn't know)
I see, thanks a lot for the feedback! That's a real hack of my LLM haha, I'll fix it within the hour!
Just tried this on my phone and had fun. Got 260/700 in Sky — somehow “peanuts” and “Michael Jackson” were my only perfect answers lol.
Thanks for trying it garah!
Very nice visual and sound experience, congrats!
Thanks a lot!! glad you enjoyed it :)
Rarest is terrible. I absolutely know that baked ube is a street food, and way less likely to be guessed than Udon, and yet I'm penalized for nobody else guessing it yet?
I didn't even play latest and oldest, because that's just a guessing game.
I'm sorry but is this not just another AI remake of a very common game type?
I recall playing an identical game a few months ago, also on the HN front page.
https://www.krillion.org/
Thanks you for the feedback, the concept doesn't come from Krillion or from me, there was already a game called Pointless on TV way before that, and Sky is straight up Family Feud, Past and Future are our own concepts And the graphics, even if the code and the implementation were obviously written with AI like 100% of people coding a big project today, they're human made
But the Abyss mode looks almost identical to krillion - from the very concept of trying to guess rare answers, to the nautical diving theme, down to the almost-pixel-art style..?
It's incredible to suggest that this wasn't based on either Krillion, or that both games were based on something else.
(To be clear, I'm not accusing you of explicitly lifting the concept - but that's the problem with using AI to generate too much, it'll obviously re-use what it's seen before.)
edit: actually, you say "A large part of the code was written with AI tools, but the artistic direction the concept and every choice design are human-made" - it's just wildly improbable for the two to look and feel so similar.
Most wonderful fact is that 55% of people dream of visiting Paris. Of course living just next to Paris puts a damper to any desire to visit
I also grew up next to Paris so I completely get it. Once you know it, it's less of a dream. Thanks a lot for your feedback :)
I love the visual aesthetic. Fun idea!
Thank you so much! The pixel art took a lot of iterations, glad it shows!
Love the visual style!
Thanks a lot! Glad you like it