ChatGPT Operator: Prompt to Autonomy · 18 min · 130 XP
Framing a question and finding primary sources
Decide what you're comparing before you collect anything.
Research goes wrong at the start far more often than in the middle. "Which library is better?" has no answer, so whatever comes back is a set of criteria someone else chose, weighted in a way you never saw.
Fix the question first: what exactly, on which criteria, where, and as of when. "Which of these two libraries has better study facilities — comparing opening hours, quiet study spaces and power sockets — in this city, as of today." Now the answer has a shape, and a missing criterion is visible rather than silently dropped.
The date matters more than it looks. Opening hours changed during the pandemic and changed back; a page written in 2021 can be accurate, well-written and useless. Recording when you checked is part of the finding, not bookkeeping.
Plan the sources before collecting them. Map each criterion to where it could be answered, and mark the ones no public page will settle — some things genuinely require phoning someone, and knowing that up front stops you accepting a weak substitute later.
Then go to primary sources. The library's own page beats a listing site that copied it two years ago. Whether you use a browsing tool or open the pages yourself, record the title, the URL and the date you looked — those three fields are what make a claim checkable by someone else, including future you.
Practice. Choose a low-stakes comparison — two public libraries' study facilities works well — and write the question with three criteria, a location and a date. Ask for a source plan mapping each criterion to where it could be answered, and flag any that need a phone call. Then locate the official page for each, recording title, URL and date checked, marking anything you couldn't find as unknown.
Loading your workspace…