Browse the deck

• Make Something Ask for five, keep the best of each Don't ask for the answer. Ask for five genuinely different attempts, then build the real one from the best parts.
• Think Better Ask for the assumptions Before you accept an AI recommendation, make it show you the load-bearing assumptions underneath.
• Think Better Ask for the missing options When you're choosing between A and B, the best move is sometimes C — and nobody in the room has said it out loud.
• Work Smarter Ask for the range, not the number Single-number estimates are confident fictions. Estimate in ranges, and see which step carries the risk.
• Do This Ask for three options, not one answer A single answer invites you to accept or reject it. Three ranked options make you compare — which is where thinking happens.
• Know the Limits Ask if this could have changed AI tools learn from material with a cutoff date. For anything that moves — prices, laws, versions, people in jobs — ask how stale the answer might be.
• Behind the Curtain Ask twice, learn something Ask the same question in a fresh conversation and compare. Agreement and disagreement are both information.
• Do This Ask what it needs from you Before a big request, flip the interview: make the AI tell you what information would improve its answer.
• Think Better Ask what usually happens Your project feels special. Projects like it have a track record. Ask about the track record.
• Know the Limits Ask where it's least sure Every answer comes out in the same confident voice. Make the uncertainty visible before you rely on it.
• Know the Limits Brief the expert, don't replace them For legal, medical, tax, and other high-stakes questions, AI's best role is making your hour with the professional dramatically more useful.
• Make Something Brief the mood, not just the specs For anything creative, the feeling you're after is a spec — usually the most important one. Say it out loud.
• Know the Limits Check the citation at the source AI tools produce references that look perfect and are sometimes wrong — right author, plausible title, no such paper.
• Behind the Curtain Confidence is a writing style AI answers sound sure of themselves because they're trained on prose that sounds sure of itself — not because they've checked.
• Human Advantage Delegate the volume, keep the judgment Sort your work into two piles: the kind that's heavy because there's a lot of it, and the kind that's heavy because it matters.
• Know the Limits Double-check the arithmetic Language models write about numbers more fluently than they compute them. The prose will be lovely; the total can still be wrong.
• Do This End with a quality pass Before you use an AI draft, send it back once: check it against your own list of what good looks like.
• Make Something Generate constraints first Before asking for ideas, make the AI interview you about the box the ideas must fit in.
• Know the Limits Give it a refusal condition Tell the model when you'd rather have 'I don't know' than a confident guess.
• Behind the Curtain It only knows what's in the conversation The model isn't quietly aware of your business, your last chat, or what you meant. The conversation is its entire world.
• Know the Limits Know what never gets pasted Decide once, in advance, what categories of information never go into a consumer AI tool — then you never have to decide under deadline.
• Work Smarter Let the goal write the agenda Start from what the meeting must produce, and work backwards to what's worth discussing.
• Think Better Make AI argue with itself After you get an answer you like, ask the model for the strongest case that it's wrong.
• Work Smarter Mine the thread for commitments A long email or chat thread is a to-do list wearing a disguise. Extract who owes what, by when.
• Do This Name the reader before you ask The same explanation lands differently for a new hire, a customer, or your CFO. Say who it's for.
• Make Something Order the ugly first draft Ask for a rough draft you intend to tear apart. Reacting is easier than creating — use that.
• Do This Paste, don't describe Stop summarizing your problem to the AI. Paste the actual email, error message, or paragraph and let it see what you see.
• Human Advantage Pick the target before you optimize AI is relentless at improving whatever you point it at. Pointing it at the right thing is still entirely your job.
• Think Better Reframe before you solve The first framing of a problem is rarely the best one. Collect three before picking a direction.
• Do This Rehearse the hard conversation Practice the raise request, the price increase, or the awkward feedback against an AI playing the other side.
• Make Something Revise in rounds, not one giant prompt Two short rounds of draft-and-feedback usually beat one elaborate mega-prompt.
• Work Smarter Run a five-minute weekly review Paste the week — what you did, what stalled — and end with three priorities you actually believe.
• Think Better Run a one-minute pre-mortem Before committing to a plan, have the AI tell the story of how it failed.
• Human Advantage Separate evidence from inference Ask for the answer sorted into what's supported, what's inferred, and what's still open.
• Do This Set the format before the content Decide what the output should look like — a table, five bullets, one paragraph — and say so up front.
• Do This Shrink it until it fits on a sticky note Summarize in halves — full page, half page, three sentences, one — until only the point is left.
• Human Advantage Some words you have to write yourself The apology, the condolence, the thank-you that matters — the value of these words is that you wrote them.
• Think Better Steelman the other side Before a disagreement, have the AI make the other position's best case — not the version of it that's easy to beat.
• Work Smarter Turn a chore into a checklist Any task you do badly every month deserves twenty minutes of checklist-building once.
• Work Smarter Turn notes into a decision Messy meeting notes become an actual decision when you force them into a fixed structure.
• Behind the Curtain Where made-up answers come from A language model on its own doesn't look things up and occasionally lie — it composes likely-sounding text, and sometimes likely-sounding is all it is.
• Work Smarter Write for the skimmer Most readers of your email will give it a few seconds. Structure for them, with the full detail underneath.

No cards match. Try a different word or category.