Playing every game of Wordle simultaneously

https://chriskw.xyz/2025/08/24/Hyper-Wordle/

by chriskw • 11 months ago

65 28 11 months ago

re• 11 months ago

the list of 2315 valid secret words
2315 was the original number of Wordle secret words. After the New York Times acquired Wordle, it was revised down to only 2309 secret words.

The NYT puzzle is no longer beholden to any specific list for solutions. There's now a larger list (of 3200 words) that Wordlebot thinks are now likely to be solutions based on NYT word frequency data, but some recent solutions (like KEFIR and LORIS) have come from outside even this set. (The NYT did however add those words into the Wordlebot dictionary just before those puzzles went live.)

12972 possible 5-letter words

Likewise, the number of legally guessable words has been increased to 14855.

chriskw3• 11 months ago

Good catch! I didn't know about this, I'll update the footnote. Anecdotally everyone I know was very surprised by KEFIR the other day.

tantalor3• 11 months ago

Because it's not a f*ckin' English word!

LPisGood• 11 months ago

Which 6 words did NYT remove?

chriskw3• 11 months ago

You can check the diff on the file wordlist_hidden here: https://github.com/alex1770/wordle/commit/85c4a7e6

jibal3• 11 months ago

agora fibre lynch pupal slave wench

cheschire• 11 months ago

I wonder if model synthesis (wave function collapse) would be more or less efficient.

https://en.m.wikipedia.org/wiki/Model_synthesis

https://www.rserra.it/solving-hardest-sudoku/

chriskw3• 11 months ago

I could see it being handy to estimate probability where each word is to better inform each guess. The tricky thing is I'm not sure if each run of wave function collapse (with randomness injected) would be an accurate sampling method for the real distribution of possible permutations. While doing this writeup I tried to find ways to analyze permutations with restriction, but it turns out most general methods are pretty intractable: https://en.wikipedia.org/wiki/Permanent_(mathematics)#Enumer...

qingcharles2• 11 months ago

ChatGPT Agent is pretty good at playing Wordle. It's banned from the official version, but it'll play the knock-off ones just fine for you.

ktallett2• 11 months ago

What's the point? You play these games to improve your own mind, or for your own enjoyment.

contravariant• 11 months ago

No you see chatgpt is doing the thinking for us now, instead of using your own brain to struggle with a puzzle that is trivial to solve for a simple algorithm you can let a language model struggle with it for you! You get all of the enjoyment from solving a puzzle with an inefficient overly complicated general purpose solver, without needing to put in any work yourself!

lemmsjid• 11 months ago

I enjoy wordle and wouldn’t automate it to replace my play. I also enjoy techie people demonstrating that you can automate X with Y tool. That’s another form of problem solving. Can’t those enjoyments exist simultaneously?

renewiltord3• 11 months ago

Same as the point in watching dog play with ball. You can also play with ball for enjoyment but sometimes the enjoyment is in seeing the limits of something else.

qingcharles2• 11 months ago

It was just a bit of fun to test the model, see what it could do :) But still, more people watch football than play football, right? ;)

oliwary• 11 months ago

Fun analysis! I run https://squareword.org, which has 5x5 letters featuring words both down and across. Guesses are applied to the horizontal words, but the vertical words can be used as support, a bit like in Sudoku.

I wonder how the analysis would work here - since the verticals form words as well I think you get a lot more information from each guess. Many players can now solve them consistently in 6 or 7 guesses.

Playing every game of Wordle simultaneously

Related Stories