Reflex Lab

1 min read Original article ↗

A miniature of the real problem: one attacker, one defender, no elixir. The project's real agent plays full matches and trains for days on a laptop CPU; this lab shows the same kind of learning, small enough to watch. The code is on GitHub.

Card roster built largely by Ambash. Built with the help of AI tools. Unofficial fan project, not affiliated with, endorsed by or sponsored by Supercell.