What good prompting technique that made a noticeable difference in the answers you get

[DiamondDallas_X]

Asking the model to list its assumptions before answering cut down on wrong guesses about my actual intent more than any other single change I've made

[Andy89]

Telling it explicitly what NOT to include cuts down on irrelevant padding more effectively than describing what I do want in detail

[CMPunk]

Asking for the answer in a specific format upfront, bullet points versus prose versus a table, gets a noticeably more useful response than leaving format unspecified

[SockPuppet]

Providing a concrete example of the output style I want works better than any amount of abstract description of my preferences

[James95]

Telling it who the audience is changes the answer more than almost any other single instruction, a technical explanation for a beginner versus an expert comes out completely differently

[TaxSeason37]

Asking for the reasoning before the final answer rather than after catches errors in the logic that I'd have otherwise missed if I'd only read the conclusion

[Zenith Theo]

Breaking a complex request into smaller sequential steps rather than one giant prompt consistently produces better results for anything with multiple moving parts

[ReacherLynx]

Asking it to identify what information it's missing before answering catches gaps I hadn't thought to specify myself