Nicholas Renotte: I Spent 6 Weeks using Reinforcement Learning to Beat This GoDot Game

Spent 6 weeks training a Stable Baselines 3 agent on a custom GoDot game—covering everything from hooking up GoDotRL to reward-shaping best practices and crafting rock-solid environments for RL. All the code (14 final versions!), logs and assets live on GitHub, with the heavy DMG files available on request. This whole thing was sparked by a mining-optimization RL demo and a desire to build an absurdly tough game.

Big thanks to Ed Beeching (GoDotRL wizard) and Brackeys (legendary tutorial inspo). Nick had a blast coding—even at 8:29 pm on a Sunday—and is off to present in Florida. If you’re nearby, give him a shout! Connect on LinkedIn, GitHub, Discord or Patreon and keep on hacking.

Watch on YouTube