Prompts that actually work.
Every prompt in here is written to be run as-is. Copy, paste your context, get a specific answer. Grouped by the kind of work you’re doing, not by 20-word category names.
- Work7Meetings, memos, difficult emails, project kickoffs.
- Code7Reviews, refactors, migrations, debugging plans.
- Research7Papers, filings, briefs, primary-source reads.
- Writing7Drafts, edits, translations, plain-language rewrites.
- Learning7Explain-like-I'm-smart on hard subjects.
- Life admin7Contracts, insurance forms, appointment prep.
- Teams4Shared rubrics, handoffs, running a small team well.
- Founders4Board memos, hiring loops, sanity checks.
Turn a difficult thread into one calmer reply.
For when the thread has gone off the rails and you need to be the person who resets it, without conceding anything real.
WORKTurn meeting notes into a decision log.
The pattern that separates a note-taker from a person the team relies on: crisp decisions, owned actions, named open questions.
WORKWrite a weekly status update people actually read.
The template your manager will forward — not because it's polished, because it's useful.
WORKTurn a messy Slack thread into a one-pager.
The move from '90 messages across three days' to 'here's a document you can share with the room that wasn't in the thread'.
WORKPrep for a hard 1:1.
Not a script — a plan. What outcome you're going for, the sentence you open with, and the two things you'll leave the room having done.
WORKBrief a collaborator in one message.
The message you send when you hand off work to someone else — with the context, the assumption they might miss, and the specific first move.
WORKKill a recurring meeting, respectfully.
The written replacement + the message that lets you cancel a standing calendar block without breaking the reason it existed.
CODEExplain a stack trace like a colleague would.
Skips the 'have you tried restarting' opener and gets to the specific failure — what broke, why, and the smallest thing that fixes it.
CODEGet a serious first-pass review before you push.
The critique you should have done to your own PR before your colleague spends their time on it. Structural issues, logic gaps, test holes.
CODERefactor without changing the public API.
The kind of refactor you can actually merge on a Tuesday — internal cleanup, external contract preserved.
CODEWalk me through code I just inherited.
The onboarding read you don't have a colleague to give you. Structure, purpose, gotchas, and the specific parts to be careful about.
CODEBootstrap a single-file CLI.
The 100-line utility you keep meaning to write. Proper argparse, real error handling, ships as one file you can pip install.
CODEWrite a migration script you can sleep on.
Idempotent, reversible, and specific about what it assumes. The script your future-oncall version won't hate you for.
CODERead a postmortem, extract what generalizes.
The pattern that lets you learn from other teams' incidents without reading 40 pages of narrative every time.
RESEARCHRead a 10-K like an analyst.
The reading order and the specific sections that carry the story. Skip the marketing letter; go straight to the footnotes that matter.
RESEARCHScaffold a literature review.
The map of the field before you write the prose — where each paper fits, what the disagreements are, and where the gap you're actually filling lives.
RESEARCHTurn a dense law into a two-page policy brief.
The version your team can actually act on — what it does, when it takes effect, who's affected, and what to do about it.
RESEARCHInterview prep from primary source.
The three questions worth asking, drawn from what they've actually written or said — not their Wikipedia page.
RESEARCHFact-check a claim, with a verdict.
Not 'here's more context' — a specific verdict on a specific claim, with the sources that would hold up to a follow-up question.
RESEARCHCompare two papers on the same question.
Where they agree, where they actually disagree (not where they use different words), and what would resolve the disagreement.
RESEARCHTrace a claim to its original source.
The pattern that separates a citation chain (secondary quoting secondary) from an actual finding, and tells you which one you have.
WRITINGCut a draft in half.
The line edit that shortens without flattening — keeps the argument and the voice, drops the parts you were only including because you wrote them.
WRITINGThree headlines, then the reasoning to choose.
Not one 'perfect' headline. Three real options at different registers, plus the reasoning to pick the one that fits your context.
WRITINGRewrite this in my actual voice.
The move that turns a competent AI draft into something you'd send. Anchor on real samples; get back your register.
WRITINGTranslate and keep the register.
The translation you actually want — where formal stays formal, dry stays dry, warm stays warm, and the target language reads native.
WRITINGCover letter without the buzzwords.
Three paragraphs, no 'passionate about', no 'excited by this opportunity'. Why you and this role fit, honestly.
WRITINGCut the adjectives, keep the meaning.
The pass that turns 'a powerful, elegant, intuitive solution' into an actual sentence about what the thing does.
WRITINGAdd a specific example.
The move that turns 'people often struggle with this' into 'here is the actual thing Chandra had to do on Tuesday'.
LEARNINGExplain like I'm smart, not five.
Assume intelligence, not knowledge. One concrete example, the intuition behind it, and the common misconception you'll immediately hit.
LEARNINGSteel-man what I disagree with.
The strongest version of the argument on the other side — so you either update your view, or hold your position with better reasons.
LEARNINGQuiz me on what I just read.
The retention move most people skip. Real questions on material you just finished — the kind you'd fail if you only skimmed.
LEARNINGWalk me through this proof.
Not 'clearly' or 'trivially'. Real motivations for each step, why this proof strategy over another, and where the difficulty actually lives.
LEARNINGExtract the real ideas from a podcast.
The three things you'd retain a week later. Guest's actual reasoning, not chapter headers.
LEARNINGExplain a diagram you can't parse.
Paste the image or describe it. Get back the plain-English walk-through that tells you what the arrows mean, what's async, and where the load actually lives.
LEARNINGTeach me a mental model.
A mental model, one concrete example, one place it fails, and one prompt-question that pulls it out of your head when you need it.
LIFE ADMINTranslate a contract clause.
What it says, what it means, what you're actually agreeing to, and the specific parts worth pushing back on.
LIFE ADMINDraft an insurance appeal.
Not a template. The specific letter that uses the insurer's own criteria language — the pattern that gets appeals reversed.
LIFE ADMINPrep for a doctor appointment.
The questions to ask, in the order that respects the doctor's time and gets you the answer to the thing you're actually worried about.
LIFE ADMINWrite a negotiation email.
The message that asks for what you want, gives the other side room to say yes, and doesn't come across as either aggressive or apologetic.
LIFE ADMINPlan a trip honestly.
Not the top-10 tourist list. The day-by-day plan that respects your constraints — jet lag, budget, what you'd actually enjoy.
LIFE ADMINDraft a complaint that gets resolved.
The specific structure that separates 'a strongly-worded email' from a message support actually escalates and acts on.
LIFE ADMINRead a lease. Flag what matters.
The clauses landlords tuck into a 12-page lease that will cost you specifically. This prompt finds them.
TEAMSOnboard a new hire in a week.
A specific day-by-day plan: which docs, which people, which shippable first task. Not 'here's the wiki, good luck'.
TEAMSRun a decision review, actually decide.
The 45-minute script that separates a review that produces a decision from one that produces a follow-up meeting.
TEAMSUnblock a project that's been 'in progress' for months.
The diagnostic that separates the reported blocker from the actual one, then names the smallest move that changes the state.
TEAMSGive peer feedback that's honest and useful.
The structure that separates 'this feedback made me want to be better' from 'thanks for the encouragement, back to work'.
FOUNDERSWrite the board update.
Numbers first, the truth about what's not working second, the one ask you need third. No burying the lede in narrative.
FOUNDERSSanity-check a hire before you offer.
The pattern that surfaces the concerns you know but haven't said out loud — the ones that will matter in six months if you ignore them now.
FOUNDERSWrite the fundraise email you've been avoiding.
The message to the investor who passed, the warm intro who ghosted, or the fund you haven't earned yet. Honest, not thirsty.
FOUNDERSPrep for a customer call you're dreading.
The three things they'll say, the specific concession you can offer, and the one thing you must not agree to under pressure.