@maharshii - i have noticed that LLMs like claude and gpt-4o work really

i have noticed that LLMs like claude and gpt-4o work really well with this prompt, it instructs them to 'contemplate' for a bit before giving the final answer.