I Spent 6 WEEKS Training AI To Beat This Level

Nick dove head-first into Reinforcement Learning over six weeks to conquer a brutal custom GoDot game, bantering with Claude along the way and building out 14 polished versions. He’s pumped about how RL can power all sorts of simulated environments—big shout-outs to Ed Beeching for GoDotRL and Brackeys for the tutorial that kicked this off.

All the code lives on GitHub (links in the post), and he can share those 167 MB env files on request. He’s off to present in Florida soon, so if you’re local, hit him up—and don’t forget to connect on LinkedIn, Facebook, GitHub, Patreon, or Discord for more RL fun!

Watch on YouTube