Comment by deprave

Apologies, I didn’t mean to imply it’s a benchmark, I just wanted to provide a reproducible example of where I see models make decisions that seem to be fine initially but might paint the software architecture into a challenging corner. I don’t expect models to read my mind, but I do see them produce a lot of verbose output, none of which is used to say “here’s a simple response to your ask, but have you also considered...”