research

Research a question

Used well, AI turns a vague question into a fast first map of the territory — what's known, what's argued, what to read next. Used carelessly, it hands you confident nonsense. The difference is whether you treat it as a research assistant you direct and check, or an oracle you trust. Here's the assistant version.

The move, in four steps

  1. Frame it out loud. Tell it the question, why you're asking, and what a useful answer would let you do. A sharper ask gets a sharper map.
  2. Ask for the shape, not just the answer. "What are the main positions on this, and who holds them?" beats "what's the answer?" — it gives you something to compare rather than swallow.
  3. Pressure-test it. Ask "what's the strongest case against that?" and "what would change your mind?" The disagreements are where the real understanding lives.
  4. Verify before you rely. Take the specific claims — names, numbers, quotes, sources — and check them against the actual source. This step is not optional.
Where it bites: it can invent sources that don't exist, misquote real ones, and state all of it with total confidence. So it's brilliant for finding threads to pull and questions to ask — and untrustworthy as a final citation. You stay the fact-checker; that's the part of the job that's yours.

Vague vs. directed

Vague
"Tell me about electric cars." → a tidy, generic essay you can't do much with.

Directed
"I'm deciding whether an EV suits me in New Zealand. Lay out the main trade-offs people argue about, the strongest case on each side, and the three facts I should verify for my own situation." → a map you can act on, with a built-in to-check list.

Carry it on

Notice the through-line: you ask for the territory and the disagreements, then keep the judgment about what's true. The same habit powers the other moves.

← Back to what you can use it for