Make sense of a lot
A long thread, a dense report, an hour of meeting notes — more than you've time to read closely. AI is genuinely good at boiling that down to its shape. The skill isn't asking for "a summary"; it's telling it what you need from the pile, and then checking it kept the bit that matters.
The move
- Give it the pile and your purpose. Paste the text and say why you're reading it: "I need to reply," "I missed the meeting," "I'm deciding X." The purpose shapes a useful summary; "summarise this" gets a shapeless one.
- Ask for structure, not a blob. "Give me the three decisions made, who owns what, and any open question" beats a paragraph. Structure is easier to scan and easier to check.
- Pull the thread you care about. Once you see the shape, go back in: "what did Sam actually agree to?" A digest is a starting point for your attention, not a replacement for it.
- Spot-check against the source. Glance back at the original for anything you'll act on. It's the cheapest insurance against a confident summary that quietly dropped the one line that counts.
Where it bites: a summary is a set of choices about what to leave out — and the model doesn't know your stakes. It can smooth over the caveat, the dissent, or the deadline that was the whole point. So you decide what counts, and you keep the original within reach.
Carry it on
Same shape as the rest: it does the volume — reading and condensing — and you keep the judgment about what actually matters. Have a look at the other moves whenever you like.