I have a feeling that after enough slop has entered the system, the AI will also have difficulty debugging/understanding it. My questions are: will the AI get

I have a feeling that after enough slop has entered the system, the AI will also have difficulty debugging/understanding it. My questions are: will the AI get to be above our level at creating grokkable source code before it comes unmanageable? And even if not: will the models' ability to understand and modify slop outpace it's ability to create it? For our jobs, I hope neither is true. But we'll see. Even in the best case we'll have a lot of cleaning up to do.