On: Learning to Reason with LLMs

[cruffle_duffle]

Okay? And? What does that have to do with anything. I thought the number one rule of these things is to not trust their output?

If you are automatically formulating some chemical based on JSON results from ChatGPT and your building blows up… that is kind of on you.