Antimatter: Mahjong Solitaire with Opposite Word Matching
The author has created a word association game where players match opposite word tiles, inspired by the Mahjong solitaire mechanic. The game is built using a large semantic graph and AI language models.
Why it matters
This project demonstrates the potential of combining AI language models and semantic knowledge graphs to create novel and engaging puzzle games.
Key Points
- 1Matching opposite words in a Mahjong solitaire-style game
- 2Leveraging a large semantic graph and AI language models to generate puzzles
- 3Exploring ways to create more interesting puzzles beyond basic word associations
- 4Potential to expand the game into a mobile experience with increased complexity
Details
The author is a fan of word association games and has created a new game called Antimatter that combines the Mahjong solitaire mechanic with matching opposite words. The game is built using a large semantic graph containing over 100 million edges, which was created through manual lexicography and inferences from various language models. The author believes that current frontier large language models (LLMs) are limited in their ability to generate truly interesting puzzles, as they tend to circle a small pool of concepts. By leveraging the semantic graph, the author has been able to create 20 algorithmically generated levels that randomly select puzzles. The front-end of the game was built using the Claude Code platform. The author is considering expanding the game into a mobile experience and increasing the complexity and challenge for players.
No comments yet
Be the first to comment