The Take
Every pick this week has something in common. None of them ask the LLM to create something from scratch.
— One hands it your last answer & says, “find the weak spots.”
— Another hands it the task you’re stuck on & says, “shrink it.”
— The 3rd hands it someone else’s writing & says, “break it down.”
The starting point is never a blank page. It’s always something that already exists.
Most people open the chat box with zero context. Write me a cover letter. Build me a business plan. Give me 10 ideas.
That’s the easy way. It’s also why the output comes back as slop.
The model works better when you hand it something real & tell it what to do with it. An answer to pressure-test, so you catch what you missed. A project to break down, so you actually start. A piece of writing to reverse-engineer, so your writing gets sharper.
Flip it.
Hand the model what you already have & tell it to double-check, simplify, or decode it.
The skill isn’t always writing better prompts. Sometimes it’s just pointing the model at something that already exists & saying “what’s wrong with this.”
You already have the material. You just didn’t think to use it as the prompt.
Seen This Week
1. Make the LLM critique its own answer
You asked a question. The model gave you an answer. But it also made choices you didn’t see. What it assumed you meant. What it filled in on its own. What it left out because you didn’t ask.
This prompt makes it go back & show you where it guessed.
Look at your last response. Find the 3 weakest assumptions in it. For each one, say what you assumed, why you assumed it, and what would change if it were wrong.
If you can't find 3 real assumptions, say so. Don't invent one to fill the count.Why it works: Most prompts ask the LLM to produce. This one tells it to check itself. It finds the gaps you didn’t think to ask about, which is the whole point of a 2nd opinion.
One gap: “Weakest” has no definition. The model picks what sounds weak, which sometimes means inventing a critique just to hit the count of 3.
Simple fix: The last part (“if you can’t find 3, say so”) already handles this. Most versions of this prompt online skip it.
Want to go deeper: After it lists the 3, paste back: “Pick the one most likely to be wrong. Be the devil’s advocate on your original answer. Be brutally honest.”
Original post by u/adCold1610. Link
2. Turn a big task into a no-brainer checklist (Pomodoro Technique)
You know what you need to do. The project is sitting there. Starting is the hard part because the whole thing feels too big. This prompt throws the project away & asks the LLM for 1 thing: a checklist so simple you can start without thinking.
I'm stuck on [TASK]. Break it into a single 25-minute starting move. Not the whole project. Just the next 25 minutes.
Give me a step-by-step checklist of what to do in those 25 minutes. Each step should be small enough that I don't have to make decisions. If something on the list needs information I don't have, mark it and skip it.Why it works: The model can’t give you a roadmap for the whole project. It’s locked into one small window. The checklist comes back small enough that not starting feels worse than starting.
One gap: Not every task has a first step you can do alone. Sometimes you’re waiting on someone else, or the real first move takes hours of uninterrupted focus. The prompt doesn’t account for roadblocks.
Simple fix: Add: “If this task can’t be started in 25 minutes alone, tell me what needs to happen first.”
Original post by u/EQ4C. Link
3. Reverse-engineer anyone’s writing
A LinkedIn post stops you mid-scroll. A tweet you’ve been trying to deconstruct. The opening paragraph of an article that hooks you before you know why.
You want to write like that, but “I liked it” doesn’t teach you anything.
This prompt takes the piece apart. It names the structure, the hook, the voice, & the close. Then it builds a reusable template & applies it to your topic in one pass.
Here's something I want to learn from: [PASTE THE WORK].
Break it down:
1. The structure. How it's organized, beat by beat.
2. The hook. What makes the opening work.
3. The voice. Sentence length, vocabulary, rhythm.
4. The close. What makes the ending land.
Then write a reusable template based on it. Then apply that template to [MY TOPIC].Why it works: The LLM doesn’t just describe what the writer did. It builds a template, then runs it on something you actually care about.
One gap: One sample isn’t enough. LLMs spot patterns easier with more examples, & the more you paste the closer the template gets to the real voice. But keep the samples from the same format. How someone writes on Twitter is different from how they write a blog post, even if it’s the same person. Tweets & articles are different voices.
Simple fix: Paste 2+ examples from the same writer. Don’t mix sources. If it’s a tweet you want to break down, grab 2 or 3 tweets.
Original post by md66358 on X. Link
The Teardown
Prepare for a hard conversation
People walk into tough conversations knowing what they want. They rarely think about what the other person wants. So when the other person gets defensive, they’re not ready. Some freeze. Some get angry. Some start backpedaling on the thing they came to say.
Asking for a raise. Confronting a roommate. Telling your partner something they don’t want to hear. The conversation goes sideways not because you said the wrong thing, but because you didn’t plan for their reaction.
This prompt tries to fix that.
Help me prepare for an important conversation.
Step 1: Ask me to share the context in 8-12 short bullets: who it's with, what happened, what I want, what I'm afraid of, and what a good outcome looks like.
Step 2: Write a clear one-paragraph opening statement in my voice using this structure: observation, impact, request. Then give me 3 questions to ask that move the conversation toward clarity, and 3 phrases to use if the other person gets defensive or vague.
Step 3: Give me a 60-minute prep plan: 3 quick exercises (each under 10 minutes) to calm my body, tighten my message, and rehearse. End with one sentence I can repeat right before I start.The useful parts are buried under a template that doesn’t belong here.
Prompt Teardown
Step 2 is the whole prompt. An opening line built on what happened, how it landed, & what you’re asking for. Questions that move toward clarity. Phrases for when the other person shuts down, gets vague, or gets aggressive. That’s specific. That’s the part people need.
Step 1 is fine. Dumping context in bullets gives the model something to work with.
Step 3 is filler. “3 quick exercises to calm my body, tighten my message, and rehearse.” The model will tell you to take deep breaths & do yoga poses (not really, but you get where I’m going). None of that helps you prepare for what your boss says when you ask for more money. Or what your partner says when you bring up something they thought was settled.
And the mantra. “One sentence I can repeat right before I start.” When was the last time you could repeat something in your head while someone else was talking? You’re either going to miss what they said or look like you checked out. Even worse, you’ve lost it.
The good parts don’t need any of this. They’re strong enough to stand alone.
Build Up
Conversations never go according to plan. That’s why you don’t want a script. You want a framework: how to open, & what to do when it goes sideways.
I need to prepare for a conversation with [WHO] about [WHAT].
Here's what happened: [CONTEXT]. What I want out of this: [GOAL].
What I'm worried about: [FEAR].
Write me an opening line: what happened, how it affected things, & what I'm asking for. Keep it under 3 sentences.
Then give me 3 questions to ask if the conversation stalls or gets vague. And 3 phrases I can use if they get defensive.
Don't give me a script. Give me a starting point & a plan for when it goes sideways.The best conversation prep is knowing how to start & having a fallback for when it doesn’t go the way you planned. Nothing more.