Though another possibility would be that since LLMs generate so much code, the LLM vulnerability discovery would just keep chugging along and we'd simply settle for the same amount of potential vulns, same relative vulnerability-exploit-patch dynamics, though higher in absolute numbers.