I enjoy Advent of Code and was inspired by it, but since it runs for a short time each year, I wanted to build something in a similar spirit that people could play year-round.
In Marches & Gnats, you do not write code in a conventional programming language. Instead, you program a single-tape Turing machine by defining its transition rules, running it, and inspecting the tape to see what happened.
There are currently 33 quests. Some are small and can be solved in ~20 minutes, while others are more involved and reward careful planning. The problems cover arithmetic, sorting, parsing text, ciphers, cellular automata, and related topics.
The game also has public leaderboards that focus on solution quality (program size and efficiency), rather than how quickly a solution was submitted. I've also experimented with AI agents that attempt to solve the same quests by directly programming Turing machines under the same constraints.
I would be very interested in feedback, especially from Advent of Code players. Happy to answer any questions!
In June, I shared Marches & Gnats (https://mng.quest), a programming puzzle game (similar to Advent of Code) where you solve challenges using a Turing machine.
Since then:
- I added 13 new quests, from arithmetic basics to Elementary Cellular Automata and Sudoku.
- Rewrote the Turing machine core in Rust, making evaluation much faster and able to handle heavier tasks.
- 102 players have joined, submitting 15000 solutions; 10 players have already solved every quest.
The hardest part turned out to be the storyline. I use ChatGPT to draft outlines. It does it quite well, but shaping them into something with real depth and atmosphere takes far more work than I expected.
Another challenge: since it's a competitive game, players quickly explored the edges of the rules. For example, submitting very long solutions that use transitions as a kind of memory. I love that kind of creativity, but it also undermines the original goal of solving a puzzle as efficiently as possible. So I've spent quite some time balancing mechanics to reward creativity without encouraging loopholes much.
The most fun part, though, is still inventing new puzzles.
I enjoy Advent of Code and was inspired by it, but since it runs for a short time each year, I wanted to build something in a similar spirit that people could play year-round.
In Marches & Gnats, you do not write code in a conventional programming language. Instead, you program a single-tape Turing machine by defining its transition rules, running it, and inspecting the tape to see what happened.
There are currently 33 quests. Some are small and can be solved in ~20 minutes, while others are more involved and reward careful planning. The problems cover arithmetic, sorting, parsing text, ciphers, cellular automata, and related topics.
The game also has public leaderboards that focus on solution quality (program size and efficiency), rather than how quickly a solution was submitted. I've also experimented with AI agents that attempt to solve the same quests by directly programming Turing machines under the same constraints.
I would be very interested in feedback, especially from Advent of Code players. Happy to answer any questions!