soupape34·vor 7 Stunden·discussInteresting thread. On the verification side — do you think formal methods or fuzzing scales better for catching the class of bugs LLM-generated code introduces?