Do you have any evidence that the bugs were due to LLM code? Or are you just speculating baselessly