Comment by throw-10-13

The human context window is the limiting factor.

That and testing/reviewing the insane amounts of ai slop this method generates.