Show HN: WordGlyph – a novel word guessing game
wordglyph.xyz
The origin of the idea came around 15 years ago. I wondered if it was possible to come up with a board game like chess but with words. The best concept I had was creating an alphabet made of segments. Each player thinks of a word. Then each player would take a turn to try to create their word stick by stick. But they eventually end up blocking each other, forcing them to come up with another word. I made the board game version and it was fun to play so then I coded it to play on FB where you could play against other people and play against the computer. But back then FB authentication kept changing, seriously like every week, which drove me crazy, constantly breaking the game. The game was not getting traction so I shut it down. Honestly, looking back it was way to complicated. Here was the instructional video I made 14 years ago. https://youtu.be/C4WbLYLTEDk?si=s0Hfn6e1Dg0lm8hf
Then when Wordle got popular I thought about my game again wondering how I could make it a single player word guessing game. But I didnt act on the idea cause I was involved several other projects. Then about a month ago there was suddenly time and I started.
The first version of the game wasn't phone friendly. But then I came up with the big matrix controlling the small ones. Also, the original scoring wasn't working well. It was very hard to score "high" (less sticks is better) because of all the random guesses were crushing the score, especially that first letter. The better I got at playing the game the more unused sticks I would have. And I thought you should get credit for that. And then it dawned on me that the unused sticks should erase the guesses and suddenly it all balanced out and felt right!
The next issue was the word list. I started with only 500 words that I asked chat to come up with. It did a great job. But then when I wanted to create all the common 6-letter words it failed. It just would not create a large list of words. At one point it literally created a csv file full of 6-letter gibberish.
At that point I realized I had to read thru all 23k 6-letter words. I already had the list of all the Scrabble dictionaries combined from when I did the project 14 years ago.
It took a week for me and some family members to go thru all the 23k words taking out obscure words. Interestingly, there are only about 4k out of the 23k that are common words.
I coded it all in css and JS. I tried to use AI to bootstrap the game from pictures of the board game. It couldn't do it. I tried to describe the game in detail but it failed to make the matrices. So I had to get the basic matrices working myself and once I showed AI my code then it finally understood what I was doing and it helped out. Claude was certainly better than openai. I will also say that currently there is no way someone who does not know how to code could have made this game just with AI. So many times it got itself stuck it a pit. Often times it would say to me "Your right the simpler approach is much better!"
The code runs on cloudflare static pages with a single worker to get the random word. Cloudlfare pages so far is a dream come true with their gift of free unlimited bandwidth.
Looking forward to hearing your feedback!