Helpful stuff, though I think the "think step by step" instruction isn't really necessary any more as long as you're using reasoning models. The only other thing I'd add is to avoid asking leading questions - you're a lot more likely to get hallucinations if you ask "What government policies led Kazakhstan to be a leader in AI?" than if you ask "Who are the countries leading in AI and what government policies helped them achieve leadership?" (Though today's models are much better than the old ones at responding to the first question by telling you that the premise is wrong.)
I turned this article into a custom promt and baked it in to the custom instructions space in personalization tab in chatgpt . Thanks for the article.
can you share?
Helpful stuff, though I think the "think step by step" instruction isn't really necessary any more as long as you're using reasoning models. The only other thing I'd add is to avoid asking leading questions - you're a lot more likely to get hallucinations if you ask "What government policies led Kazakhstan to be a leader in AI?" than if you ask "Who are the countries leading in AI and what government policies helped them achieve leadership?" (Though today's models are much better than the old ones at responding to the first question by telling you that the premise is wrong.)