Yes but it takes much longer to trace them. Because the LLM code almost always gravitates toward data blobs and highly dynamic objects and spaghetti that takes

Yes but it takes much longer to trace them. Because the LLM code almost always gravitates toward data blobs and highly dynamic objects and spaghetti that takes a ton of cognitive load to understand what their failure modes are. Even when it does document them.