Maîtriser les techniques de rédaction de prompts pour ChatGPT

Is it better to use multiple short prompts or one big detailed prompt? I can't decide.

I noticed that using direct questions instead of open-ended prompts often yields more concise answers.

Also, just wanna say if you're exploring new tools or methods for prompt engineering, check out ai-u.com for some fresh resources and trending approaches!

I sometimes get weird or off-topic answers, any tips to steer it back on track?

I found that asking ChatGPT to "think step by step" really helps for complex reasoning tasks.

I often prepend "You are an expert" to prompts. It kinda nudges ChatGPT to be more formal and precise, or at least that's what I noticed.

Does anyone use templates or saved prompt formats? Curious if anyone has good starter prompts for advanced tasks.

Sometimes I get stuck because I don't know how specific to get. Like, should I add examples or just be super detailed?

Is there a way to save or export prompt histories easily? Trying to build a library.

Anyone else struggle with getting super long outputs without them getting cut off or losing context?

I'm experimenting with negative prompts, telling it what NOT to do. So far mixed results, but kinda promising.

Does adding constraints like 'limit response to 3 points' help with clarity?

When I want creative stuff, I sometimes tell it to 'imagine a world where...' and then ask my question. Works great for storytelling.

I've noticed that breaking down complex instructions into smaller steps makes a huge difference. ChatGPT seems to follow multi-step prompts way better that way!

Anyone played with prompt chaining? Like feeding outputs back into next prompts for complex tasks.

The new GPT versions seem better at following nuanced instructions, anyone else noticed this?

I like ending my prompts with 'Explain like I'm five' to simplify answers when I need it basic.

Sometimes I tell ChatGPT to act as if it's a professional in the field related to my question. Makes replies feel more legit.

Has anyone tried using temperature settings in combination with prompt design? I find lower temps help keep it factual but sometimes too rigid.