Ben Pace comments on Computational complexity of RL with traps

[Ben Pace]

I am curious about what you were thinking (Seems potentially useful to me/​someone else who might try the same idea or have a similar misunderstanding)