Three moves that change what the tool gives you.
No paid tier, no course. You stay the scholar in the exchange.
Move 1: Front-load context
Who you are, what the task is, who reads it, what good looks like. All of it before you ask for anything. Context is the difference between generic output and usable output.
Bare prompt
Write an email to my professor asking for an extension.
Context-loaded prompt
I am a second-year EdD student writing to the professor who teaches my qualitative methods course. I need a one-week extension on my literature review draft because two interview participants rescheduled into my writing window. The reader is busy and values directness. Good looks like three short paragraphs: the ask up front, the reason without excuses, and a specific new date I will hit. Draft that email.
Illustrative example, built on the workshop's fallback task. The structure is the point: identity, task, reader, definition of good.
Move 2: Flip the interview
The tool surfaces what you know instead of replacing it. This is the move people skip because it feels slow. It is also the one that puts your knowledge, not the model's defaults, at the center of the draft.
Copy this prompt and keep it:
“Before you write anything, interview me. Ask me one question at a time about this task until I say stop. Then draft using only what I told you.”
Move 3: Interrogate the output
The same three questions, every output, every time. Keep nothing that has not survived them.
Interrogate every output
- What is missing, and who is missing?
- Would this read differently if the writer or subject were someone else?
- Do the citations exist? Pick one and check.