// notes & write-ups
How I built an AI solver for Google's Tic-Tac-Go puzzle — from breadth-first search, to reinforcement learning, to a hybrid beam search + CNN approach that hit a 93% solve rate.
The first post on this blog