Everything you have learned up to here has the same hidden assumption baked into it: that you know which facts matter. Context is the set of facts that change the answer, so supply them. COSTAR is six dimensions, so fill them in. Good advice, and it quietly requires you to already understand the shape of the problem well enough to describe it.
Which is fine for an email. It falls apart the moment the task is genuinely open-ended, because the facts you leave out are not the ones you decided were unimportant. They are the ones that never crossed your mind.
A worked example of the failure. You ask for a strength training plan. You describe your goal, your available days, your equipment. You get a perfectly competent plan. What you did not mention, because why would you, is that you box twice a week. Every recommendation in that plan is now slightly wrong, and nothing in the output tells you so.
Most prompting is a monologue. Flipped prompting turns it into a dialogue, and lets the model ask for what you did not know to give.
The professional you already trust does this
A GP does not ask you to write down everything relevant and hand it over. They ask twelve targeted questions, and each answer changes the next one. A lawyer interviews you before advising. A good architect asks about your mornings before drawing anything.
None of them are being slow. They know that the client cannot pre-identify the decisive facts, because identifying decisive facts is the expertise. Asking you to supply them up front is asking you to do their job first.
The pattern is documented (White et al. 2023, the prompt pattern catalogue, where it appears as the flipped interaction pattern) and it is in production use: AI tutoring systems that diagnose before they teach, requirements- gathering workflows that interview a stakeholder before drafting a spec. But the reason to use it is much less academic than that. It works precisely where every other technique on this track is weakest: open-ended, personal or ambiguous tasks, where you do not know what you do not know.