// the find
Code-Bullet/Google-Chrome-Dino-Game-AI
A NEAT (NeuroEvolution of Augmenting Topologies) implementation that trains an AI to play the Chrome offline dino game, built as a companion to a YouTube video. Written in Processing, so it runs as a sketch rather than a proper application. Aimed at people learning neuroevolution through a familiar, visual example.
NEAT from scratch in ~10 files is a reasonable teaching artifact — speciation, connection genes, node genes, and population management are all separated into their own files. The visualization controls (G to replay good generations, N to hide for speed) make watching evolution actually useful for learning. Processing keeps the barrier to entry low; download and run, no build toolchain.
Last touched in 2023 and the README is four sentences pointing to a YouTube video — there is no explanation of the NEAT parameters, fitness function, or how inputs are wired to the network, so reading the code is the only documentation. Processing is a dead end for anyone who wants to extend this into something real; no Python/JS port means you can't drop this into a web demo or a standard ML pipeline. No tests, no save/load for trained genomes, so every run starts from zero. The repo exists to support a video, not as a reusable library, which limits how much you can actually build on it.