Devlog #05 - Beginning on the AI

With the desktop game now complete, I started working on the machine learning side of the project. This update lays the groundwork for eventually training an AI to play Tractor.

The latest version is available on the GitHub Releases page.

AI Framework

I set up the initial machine learning framework that future models will use for training and gameplay. While the AI isn’t making intelligent decisions yet, the project now has the base structure needed to begin
testing different approaches.

Bot Support

The game now supports AI-controlled players. For now, they simply make random legal moves and are not in the main release, but this provides a baseline that can later be replaced with a trained model.

Other Changes

• Added the initial training infrastructure.
• Refactored parts of the codebase to better support AI integration.
• Continued organizing the project for future machine learning development.

Future Plans

• Create a basic hand-written bot for the AI to mimic.
• Train the first ML models to mimic the hand-made bot.
• Replace the random bot with the learned model.
• Explore self-play to improve the AI beyond imitation.