I just launched Explainle (https://explainle.org), my latest word game. Every day, you have to explain 5 words without using forbidden words. An LLM tries to guess which word you are referring to. The fewer attempts you need, the more points you score!
This is the first game I've built with Claude Code, which I have found to be very powerful indeed. Also my first game to use LLMs as part of the game - quite intrigued to see how it will scale haha. The frontend is built using Vue.js, the backend is pocketbase.
Here is my score for the day:
Explainle #24
Total Points: 5000 (Top Score)
Word 1: (1000pt)
Word 2: (1000pt)
Word 3: (1000pt)
Word 4: (1000pt)
Word 5: (1000pt)
Sadly I seem to have kindles which suffer from increasing numbers of stuck pixels after 3-5 years.
I think I'm on my third now, and while I kinda begrudge buying new ones the paper-white I've got now, with backlight, is such a pleasure to use I can almost forgive it.
Interesting - a lot of problems I face seem to fit into this general framework, and I feel like I keep reinventing the wheel. How is conlict resolution handled?
> If you’ve ever tried to construct a crossword, you’ll find that the framing of a crossword grid under square theory feels right. When you’re nearing the end of the grid-filling process, finding valid crossings of words to fill that final corner of a grid, there’s a satisfying “clicking” feeling—a sense of magic—when it all fits together, analogous to the wrapping-around feeling of completing the square.
If you enjoy this feeling, I think you would like my word game https://spaceword.org. The goal is to arrange 21 letters in a square that is as tight as possible. No one has achieved a "perfect" pattern yet, but people are very close, often leaving only 3 spaces blank!
Fun game! Though I dispute that people are "very close" to achieving a perfect pattern.
To get a "perfect" pattern you'd need to find three 7 letter words that can stack on rows adjacent to each other to form a 3 letter word in each column. Such arrangements do exist, for example:
o p e r a t e
a r r o w e d
r e s e n d s
but they are very rare - I estimate something on the order of 0.002% of combinations of three 7-letter words have any valid arrangements. Assuming that you're using standard ETAOIN letter frequencies, the typical bag of 21 letters will usually have just a handful of combinations of three 7-letter words so a given puzzle has a << 0.1% chance of having a perfect solution.
But there are 12,000x more ways to rearrange 21 tiles within an 8x3 grid, and the word choices are more forgiving as well (if you draw 7 letters from the etaoin frequency distribution, those 7 letters in order are much more likely to form a 3 letter word followed by a 4 letter word than they are to form a 7 letter word). Pretty much every puzzle should have at least some solutions fitting within an 8x3.
Additional note: 3 blank spaces is the best non-perfect arrangement, since the grid is only 10 tiles wide. One blank space could only be achieved by a single 23-letter-long word, and two blank spaces could only be achieved by a 10 letter word next to an 11 letter word, and an 11 letter word would not fit inside the 10x10 grid.
Glad you like it! :) And thank you for your comments, super interesting! Excellent point about the rarity of the perfect arrangement. Perhaps I should throw in a few lettersets that do have a solution, I am intrigued to see if people would discover it.
My other game, https://squareword.org focuses exclusively on perfect 5x5 squares, but here the goal is to uncover it wordle-style rather than arranging it from scratch. There are surprisingly few combinations that have ten unique, common words in a 5x5 letter square!
I struggled on a few days' puzzles under the assumption that there _was_ a perfect solution possible -- It may be worth noting in the "help" that not all lettersets can be solved perfectly.
Puzzles that are always open is exactly what I mean, yeah.
Some "daily" games call this kind of generated puzzle a "practice" mode. But whenever I encounter a daily game, I go straight for that mode, which is what most games would just present as the game itself.
My favorite sideprojects are daily games. One I am currently enjoying building is VideoPuzzle: https://videopuzzle.org/ where you have to unscramble a video split into 4x4 tiles.
We are up to almost 200 puzzles, with around 700 players per day. I've become much better at finding videos that work well as puzzles and am working on adding small quality of life updates.
What about insert and delete operations however? Isn't there a risk of there being too few tokens to properly finish the code in-between the "final" tokens?
I think the change here will be something we've seen with the other modalities. Text was interestingly syntactically correct but nonsense sentences. Then paragraphs but the end of the article would go off the rails. Then the article. Now it's that the creativity of the children's story in question.
Pictures were awful fever dreams filled with eyes but you could kind of see a dog. Then you could see what it was, then decent
Videos were fun that they kind of worked, then surprising it took a few seconds for the panda to turn into spaghetti, then it kept the general style for a decent time.
I see this moving towards the creativity being the major thing, or it having a few general styles (softly lit background for example).
This has mostly all shifted in a very short space of time and as someone who put RBMs on GPUs possibly for the first time (I'm gonna claim it) this is absolutely wild.
Had I seen some of this, say, 6 months ago I'd not have guessed at all bits weren't real.
Last night my girlfriend asked me why I kept watching the same bland sounding videos again and again. She came over and watched for a bit, gave a sort of confused laugh of solidarity to something, like "Uhh, why is he so into this? But, ok, I guess..." and then walked away.
It wasn't until I was able to get my jaw off the ground that I told her it was AI. No, not AI like special effects, completely AI.
A very important demo video I also found on reddit is this one [0] that's a fairly generic series of action scenes of a raid leading to a gun fight. The individual scenes are look mostly fine, notable exceptions being the muzzle flashes and nonsense guns in a few shots, but the connecting flow is nonsense if you look at it even a little. It has some of the consistency issues that are a bit of a halmark of AI videos, the interior size and layout of rooms and vehicles morphs and shifts from 'shot' to 'shot', they get out of the vehicle twice, etc. The wheels really come off in the more actiony scene though with the pace feeling very plodding for what would be an intense scene in even a moderately competitent human editor. Also some of the 'cops' wind up shooting each other in one scene which was a funny mistake.
Looks like AI crossed a line. At the very least, one person can do long form documentaries from their basement using VEO 3. There is no need for camera shoots. Yikes.
This reminds me of Pixar's video of an animated lamp 40 years ago. I remember that within 5 years Toy Story came out and changed everything on how animated films were made. Looks to me like we are on our way to doing the same thing with realistic movies.
Good point, how about something history related? Ken burns does something similar with photos. I've also seen animation used in documentaries. So use AI instead. How about a mock documentary? Spinal Tap comes to mind.
It’s true. YTMND nailed the TikTok / Vine format like 12 years ahead of its time. If only they’d “pivoted to mobile” and added more ease of use creation tools they may have stayed relevant.
Fantastic to see it's getting updated! I am a big fan of litestream, have been using it for a while together with pocketbase. It's like a cheat code for a cheap, reliable and safe backend.
reply