Show HN: Jev Cards, Can you beat a vision model at a picture quiz? (jevcards.com)

72 points by totisjosema 4 days ago

I built a quiz game around TypeSafe's new Jev model, which has vision capabilities. Jev looks at a picture and a question and picks an answer (or taps a place) in about a quarter of a second.

I've always loved quiz games and thought this would be a fun way to see what it can do and how fast it is. I made the cards from scratch, and Jev gets exactly the picture you see. More cards will come over time.

Play against Jev on your own (there's a leaderboard for each mode), or start a live match and challenge your friends, with Jev as one more player.

The code is open source, so feel free to build your own game or make more cards that I can publish.

Play: https://jevcards.com Code: https://github.com/Jose-Sabater/jevcards (the newest cards stay private for a while)

Disclaimer: I ran Jev through both TypeSafe and Opper, and made the cards through Opper's vision models and Claude Code. I work at Opper, but jevcards is open source and free to play.

TZubiri 4 days ago
Couple of bugs:
1- I got 7/10, jev got 6/10, and it says Jev won, presumably because I took 100 seconds, but Jev took 60? Funny what that reveals about author and AI-optimism in general
2- The chess board challenge isn't clear on how it should take the answer, I tried dragging a piece but that didn't work, then I tried touching the piece I wanted to move, but it marked the answer wrong, I apparently had to touch the square I needed to move the piece to.
- totisjosema 4 days ago
  Jev did not take 60 seconds for sure, with answers in 200-300 ms averages, more like 2 seconds.
  You get very high scores if you are fast and accurate (which Jev is most of the time), and you get punished hard (15s) if you are wrong. The game balances speed and accuracy.
  Will try to improve the cards that didnt work for you, thanks for the feedback!
  - TZubiri 4 days ago
    It's technically subjective, but to me speed is a tiebreaker, it goes after the decimal point. Being right is more important than being fast.
    There's notably no button for "I don't know", but if there were, a human would maybe get 7 rights and 3 I don't know. While Jev would probably get 6 rights and 4 wrongs. I get that it's a game, but it's also not.
    I think great minds make great efforts to not be wrong, someone that is never wrong is in a different category completely than something that answers fast, or even that answers more correct questions.
    totisjosema 1 day ago
    There are 3 game modes, and each tests different things, (and yes game 1 is almost impossible for a human). However it feels like you would do great in streak, where correct answers is all that matters.
    Also incorporated your feedback on versus mode :) Most accurate always wins (the systems point accounts for this)
- ajjenkins 4 days ago
  I was confused by the chess board question too. But I do think the question is worded correctly—it’s just a little unintuitive.
joshstrange 4 days ago
I beat Jev and only failed on the dart board, I have zero idea how to read a dart board.
- totisjosema 4 days ago
  Well, good that you beat Jev anyway , not many do :)
nickabe 4 days ago
You might want to adjust the scoring.
A simple system would be something like: - accuracy is most important - if tied on accuracy (ie both got 8/10) then tie break is done on speed
In general, across a lot of games - winning is a big determinant in who comes back. Right now it might be very difficult to beat JEV with the way your scoring is setup.
For my game in particular it's like a 47% chance that a winning player returns versus just 10% for non-winners.
- totisjosema 4 days ago
  Each game mode tests different things. Sprint is all about speed, it will be hard to beat JEV if he doesnt fail but one failure adds 15s. So you have a chance (in fact JEV gets around 80% right)
  Versus Measures accuracy but under time pressure, if you score more than Jev you will win.
  And streak is also accuracy.
  This quiz game, is supposed to visualize how good and bad these type of models are. They can be very fast but they can also fail 20% of the times. As a human, can you beat that?
nextaccountic 4 days ago
Could you add an option to remove musical theory questions? I mean, I know Jev can ace those as well, but I can only click at random, that's unfun
- ajjenkins 4 days ago
  In my opinion the the point of the site is that Jev has knowledge on all topics that makes it hard to beat. It might not be as smart as you, but it probably has a greater breadth of knowledge.
  - nextaccountic 4 days ago
    Yeah but this is also a fun, addictive game and I was invested into beating Jev, but every run where musical theory appeared had to be aborted
- totisjosema 4 days ago
  just play again :), but will consider adding whitelist/blacklist of categories
  - folmar 4 days ago
    also less USification would help, it's not common knowledge how US birds or bridges or hats look like
    totisjosema 4 days ago
    will add more cards, thanks for the feedback
  - hluska 4 days ago
    Considering this whole thing is AI slop, you could be a lot more polite when people make feature requests.
    totisjosema 4 days ago
    sorry you saw it that way, was just kidding, there are many categories unlucky of him that the one he doesnt like landed on his deck Anyway the feature he requested was literally on the app 5 minutes after he asked :)
ajjenkins 4 days ago
This is fun, and a good demonstration of Jev’s capabilities.
It’s obviously much faster than a human, but it’s interesting that it only got 7/10 right on both rounds (same as me). I’m wondering what Jev could be trusted to do reliably given these results.
I guess it wasn’t optimized for these types of trivia questions. Or maybe the image capabilities are too new and need refinement.
- totisjosema 4 days ago
  Thanks for the feedback, and you got the purpose of this little app exactly right. Each game mode tests different things. And it shows the strengths and weaknesses of this model. In streak almost all humans will win, but in sprint very few, however, very good humans will still beat the machine.
  Jev and decision models, can make decisions very fast, in this quiz with around 80% accuracy, each use case will need to define if its good enough or not.
jrflo 4 days ago
Very cool demo! Although I feel that incorrect answers should be penalized more. For instance- I got 100% in about 45 seconds on the sprint. Then, I clicked through as fast as I can (only limited by finding the right targets) and got a score of 42s, technically better than my first score but I only got 20% right.
- totisjosema 4 days ago
  Thanks for the feedback! And now that you showed everyone a cheat sheet we will get many high scores, so lets increase the penalty of wrong answers and reset the leaderboard! Thanks a lot for the insights
totisjosema 3 days ago
A quick update, I took some QoL feedback received in comments and also added 300 new cards so play stays interesting.
Also Ben... Whomever you are I know you cheated, and... will allow it for now ;)
arjie 4 days ago
Would prefer if the audio button was at the top right of the page etc.
- totisjosema 4 days ago
  Will take a look, should already be there during a game though
cpuhalladmin 2 days ago
Fun! Made me realise i need to brush up on geography. Site is quite smooth games are fun.
- totisjosema 2 days ago
  Thanks!
podnami 4 days ago
This was pure fun for 62s until I found out how lousy I am
- totisjosema 4 days ago
  Jev has been humbling many today (including me) Still fun to play tho, invite some games and play with them.
geooff_ 4 days ago
I "Lost" to Jev while getting 10/10 questions and it got 8/10. Interesting scoring?
- drivers99 4 days ago
  I wonder if they fixed it, or if it depends on the game type. In Versus mode, I got 10/10, Jev got 7/10, but the points were closer: 1201 to 1034 so I assume time is also part of the calculation. One of mine was 9.6 seconds vs 0.3 seconds. It wasn't always clear where you were supposed to click, like on a graph, the choices were buttons at the bottom of the screen.
  - totisjosema 4 days ago
    Its a combination of time and correct/wrong answer. But the game had not been played by these many people before so it can be that the scoring mechanism is not fully calibrated. Thanks for trying it out!
- TZubiri 4 days ago
  Too late, we've already replaced online image and video moderation with Jev, get with the times.
sara_kumodeck 4 days ago
Having the model as just one more players in a live match with friends is a fun way to show off tis speed
- totisjosema 3 days ago
  Yeah, the solo play is fun, but with friends its much better
AuthAuth 4 days ago
I went in thinking I was going to cook Jev and then I get asked things like "pick the ginko leaf". Good site
- totisjosema 4 days ago
  haha, wait until you get asked which is a dolphin
darepublic 4 days ago
Tbf jev should also be required to use vision on the UI to decide where to click to proceed with the quiz
darepublic 4 days ago
Nice site. I assume vibecoded but it was tasteful and performant on mobile. Thanks for open sourcing it
- totisjosema 4 days ago
  Thanks! Yeah built with claude code.

Showing 39 of 54 comments. Full thread on Hacker News.