One of the weakest prompts you can provide to an LLM is:

“What do you think of this?”

Instead, do something like this:

“Review this argument like a skeptical technical reader. Identify the three weakest assumptions, explain why each is vulnerable, and suggest how to strengthen them without changing the core thesis.”

The same model will provide responses dramatically different in each case.

Remember: no good model can overcome a bad prompt.