Comment by lucasfin000

This is definitely the right question. A list of failures without any baseline won't tell you anything. You would need the same exercise for human-written code at a comparable scale before drawing any conclusions at all. Without it, it's just confirmation bias.