If you are checking every line written by an LLM , what would you estimate is the probability it would ever do this?? I look at most code still and I have a high level of trust that it would never do anything remotely similar to this.