soupape34·8 時間前·議論Interesting thread. On the verification side — do you think formal methods or fuzzing scales better for catching the class of bugs LLM-generated code introduces?