Scout
Map a topic, surface vocabulary, and show what you need to learn next.
Practical guide / 01
Good results rarely come from a magic sentence. They come from a clear brief, useful context, and a few rounds of honest revision.
Start with the workflowA repeatable method
Use the model as a collaborator with incomplete context—not as a mind reader or an authority.
Say what you are trying to achieve, who it is for, and what a useful result looks like.
Paste or attach the source you want it to use. Separate facts, constraints, and examples.
Specify format, length, tone, and what should be left out. Ask for questions if context is missing.
Do not restart after an imperfect answer. Point to what worked, what did not, and what must change.
Check consequential facts, calculations, citations, and recommendations outside the chat.
Prompt lab
Adapt the brackets. The structure matters more than the exact wording.
I need to [goal]. The result is for [audience]. Use this context: [context]. Respect these constraints: [constraints]. Return [format]. Before you begin, ask up to three questions that would materially improve the result.
Act as a rigorous editor. First identify the three biggest weaknesses in this draft. Then revise it while preserving my meaning and voice. Explain the most important changes briefly. Draft: [paste text]
Audit your previous answer. Separate verified facts, reasonable inferences, and uncertain claims. Flag anything that needs an external source, and suggest how I can verify it.
Common work
Map a topic, surface vocabulary, and show what you need to learn next.
Clarify structure, expose weak logic, and create alternatives without erasing your voice.
Explain at your level, test understanding, and adjust when an explanation does not land.
Rehearse an interview, objection, meeting, or difficult conversation before it matters.
Remove personal, confidential, regulated, or client information unless you know your organization’s approved setup and policy.
A polished answer can still be wrong. Ask for uncertainty, then verify claims where error has a real cost.
Let ChatGPT help you prepare a decision. Do not let it quietly become the person making one.
Next field guide