I think indie games could be a really good use case for coding AIs. Low stakes, fun-oriented, sounds like a match.
The first commit[0] seems to have a lot of code, but no `PROMPTS.md` yet.
For example, `EnergySystem.ts` is already present on this first commit, but later appears in the `PROMPTS.md` in a way that suggests it was made from scratch by the AI.
Can you elaborate a bit more on this part of the repository history?
[0]: https://github.com/maciej-trebacz/tower-of-time-game/commit/...
Because this was a game jam entry with one week deadline I was going pretty fast and didn't bother to use source control for the first 2-3 days of work, hence the huge initial commit. I also weren't writing down prompts as I went, only after the game was finished I went back in my chat history in the tools I used and copied all the prompts to the `PROMPTS.md` file.
If you want to follow the history of this project as it was created the best way would be to read the prompts file from top to bottom. For example the EnergySystem.ts file was created right after I was done with enemy pathfinding, spawing and tower shooting and it was created from scratch by the AI using the prompt "I want to implement an Energy subsystem where..."
You can easily let the LLM take over committing. "create atomic conventional commits until the git workspace is clean".
I have a Claude Code session open just for commiting the state once I have a piece of work complete and working.
Thanks for clarifying!
Also, good catch in using the chat history to reconstruct the first phases of work.
I believe it can be a fun experiment for others to try to reproduce it from scratch using the prompts and image assets only.