An "interesting" coding challenge

I have an interesting coding challenge (you could probably tell from the title):
Create a neural network capable of learning how to play Legend of Zelda.

The issue names no repository files, tests, or entry points. Start by clarifying the intended Zelda game, neural-network approach, available game interface, and success criteria; the work is done only when those constraints and a measurable learning outcome are defined.

索引モデルが issue の本文から書いたものです。