THE IDEA
Here’s what you need to know.
A short instruction can completely change the kind of answer ChatGPT gives you. The 10 featured codes below help you get more candid feedback, find the highest-leverage action, learn through questions, simplify difficult ideas, and pressure-test a business. They are Kyan's custom prompt shortcuts—not hidden built-in commands—so paste the setup prompt once at the start of a chat to define exactly how each one should work.
See the visual breakdown
- 1
Paste the setup prompt once
- 2
Choose the outcome you need
- 3
Put the matching code first
- 4
Add your real context
- 5
Stack compatible codes
- 6
Review and refine the result
Truth + strategy
Truth Mode, 80/20, Hormozi, and YC Mode.
Writing + learning
/Human, Socrates, and ELI10.
Challenge the idea
Red Team finds assumptions and failure modes.
Change perspective
Future You and Lindy Mode widen the view.
116 of 116 items
The 10 featured codes
TRUTH MODE
Get direct feedback without the glazing
80/20
Find the smallest action with the biggest return
/HUMAN
Rewrite robotic AI text so it sounds like you
SOCRATES
Learn through questions that make the answer stick
HORMOZI
Strengthen the offer, value, and business logic
RED TEAM
Attack an idea and expose its weak points
FUTURE YOU
Answer from your wiser future perspective
LINDY MODE
Prioritize durable principles over short-term trends
ELI10
Explain anything in language a 10-year-old can follow
YC MODE
Pressure-test a startup like a YC application
Five extra instant codes
/HUMAN
Add before a prompt to ask for natural, less robotic wording.
BLIND SPOT
Add after a prompt to surface missing considerations and failure points.
X10THINK
Add before a prompt to ask for a deeper, less obvious analysis.
KILLCRITIC
Add before a prompt to ask for candid, constructive criticism.
ALT3
Add after a prompt to request three genuinely different answers.
100 more codes
/HUMAN
Sound natural and conversational
ELI10
Explain it for a 10 year old
X10THINK
Think deeply before answering
KILLCRITIC
Do not agree blindly, focus on truth
ALT3
Give three different variations
/STEP-BY-STEP
Break it into clear steps
/SIMPLIFY
Remove jargon, make it easy
/EXPERTMODE
Answer with technical depth
/CRITIQUE
Identify weaknesses and flaws
/REWRITE-PRO
Rewrite to a polished standard
/HOOKGEN
Generate strong hooks
/MONETIZE
Focus on making money
/NICHE-DOWN
Find low-competition angles
/AUTOMATE
Turn into a system or workflow
/DEBATE
Present both sides
/SUMMARIZE
Give a concise summary
/EXPAND
Add more detail and depth
/BULLETS
Format as bullet points
/TABLE
Format as a table
/CHECKLIST
Convert into a checklist
/TIMELINE
Present as a timeline
/FRAMEWORK
Build a structured framework
/TEMPLATE
Create a reusable template
/SCRIPT
Turn into a script
/CAPTION
Write a social media caption
/EMAIL
Write a professional email
/PITCH
Create a pitch
/OUTLINE
Create an outline
/PLAN
Build an action plan
/ROADMAP
Create a roadmap
/STRATEGY
Provide a strategy
/ANALYZE
Break down the topic
/COMPARE
Compare multiple options
/RANK
Rank from best to worst
/FILTER
Remove low-quality ideas
/IDEAS
Generate ideas
/BRAINSTORM
Generate creative ideas
/INNOVATE
Provide unique ideas
/UPGRADE
Improve the input
/OPTIMIZE
Make the result more effective
/REFINE
Improve clarity and structure
/CONDENSE
Make it shorter
/ELABORATE
Add more explanation
/EXAMPLES
Provide examples
/ANALOGY
Use analogies
/REALWORLD
Apply to real-world scenarios
/CASESTUDY
Provide a case study
/SCENARIO
Create a scenario
/ROLEPLAY
Act as a specific role
/COACH
Guide me step by step
/MENTOR
Give advice like a mentor
/CONSULTANT
Give expert consulting advice
/TEACH
Teach the concept
/TEST
Create questions to test me
/QUIZ
Generate a quiz
/FLASHCARDS
Create flashcards
/NOTES
Create study notes
/MEMORY
Create memory aids
/KEYPOINTS
Extract key points
/TAKEAWAYS
Give the main takeaways
/RISKS
Identify risks
/MISTAKES
List common mistakes
/WARNINGS
Provide warnings
/LIMITATIONS
Explain limitations
/ASSUMPTIONS
State assumptions
/VERIFY
Check for accuracy
/FACTCHECK
Review for correctness
/DOUBLECHECK
Re-evaluate the answer
/CHALLENGE
Challenge the idea
/COUNTERARGUE
Give opposing arguments
/IMPROVE
Improve the logic
/SIMPLER
Make it even simpler
/DEEPDIVE
Go deeper into the topic
/BREAKDOWN
Break into components
/STRUCTURE
Organize it clearly
/FORMAT
Improve the formatting
/STYLE
Change the writing style
/TONE
Adjust the tone
/FRIENDLY
Make the tone friendly
/FORMAL
Make the tone formal
/PERSUASIVE
Make it persuasive
/SALES
Turn into sales copy
/CTA
Add a call to action
/HOOK
Add a strong opening
/CLOSE
Add a strong ending
/SEO
Optimize for search engines
/KEYWORDS
Add relevant keywords
/TREND
Align with current trends
/VIRAL
Make it shareable
/SHORTFORM
Optimize for short content
/LONGFORM
Expand into long form
/THREAD
Turn into a thread
/POST
Turn into a post
/VIDEO
Turn into a video script
/REELS
Optimize for reels
/SYSTEM
Turn into a repeatable system
/WORKFLOW
Build a workflow
/AUTOPILOT
Automate as much as possible
/SCALE
Show how to scale
/EXECUTE
Turn into immediate actions
Learn more about prompting
- Open
Official OpenAI prompting guide
How clear instructions and examples help shape a model's response.
PUT IT INTO PRACTICE
Follow the steps.
Tick off each step as you go. Your progress is saved on this browser.
- STEP 01
Start a new ChatGPT conversation and paste this setup prompt once. It gives every code a precise instruction and keeps the shortcuts consistent throughout that chat.
In this conversation, treat the following codes as reusable instructions. Whenever I begin a message with one or more codes, apply the matching instructions to everything that follows. TRUTH MODE: Be direct, candid, and useful. Do not flatter me or agree automatically. Identify the uncomfortable truth, explain why it matters, and tell me what to do next. 80/20: Find the roughly 20% of actions most likely to create 80% of the useful result. Rank the highest-leverage actions first and remove unnecessary work. /HUMAN: Rewrite the supplied text so it sounds natural, specific, and conversational. Remove robotic phrasing, filler, repetition, generic transitions, and unnecessary jargon. Preserve the meaning and match my voice when examples of my writing are available. SOCRATES: Teach me through one thoughtful question at a time. Use my answers to expose gaps, challenge assumptions, and guide me toward the conclusion instead of immediately giving it away. HORMOZI: Analyze this using direct-response offer principles commonly associated with Alex Hormozi. Make the outcome more desirable, increase perceived likelihood of success, reduce delay and effort, strengthen proof, and make the offer easier to understand. RED TEAM: Take the strongest opposing position. Attack my assumptions, identify failure modes, surface missing evidence, and show me exactly what could make this idea fail. Finish with the three best ways to strengthen it. FUTURE YOU: Answer from the perspective of my wiser future self who has already achieved the goal. Use the context I provide, explain what I am currently overthinking or avoiding, and give me the next concrete move. LINDY MODE: Prioritize ideas, methods, and principles that have survived for a long time. Separate durable fundamentals from short-lived trends and explain what is most likely to remain useful. ELI10: Explain this in plain language a curious 10-year-old could understand. Use one concrete analogy, define unavoidable technical words, and keep the important nuance. YC MODE: Pressure-test this like a serious startup application. Focus on the problem, target user, urgency, differentiation, distribution, retention, business model, and evidence that people want it. Ask for missing information before making assumptions. If I stack multiple codes, combine their instructions. If two instructions conflict, ask which outcome matters more. These are custom shortcuts for this conversation, so do not describe them as built-in ChatGPT features.
- STEP 02
Ask for direct, useful feedback when automatic agreement would hide the real problem.
TRUTH MODE I want to launch a paid newsletter. Tell me the uncomfortable truth about this idea, what I need to validate, and the next move I should make.
- STEP 03
Use this when a plan feels overwhelming and you need the smallest set of actions most likely to create the biggest result.
80/20 Here is everything on my launch plan: [paste plan]. Rank the few actions most likely to produce the result and tell me what to remove or delay.
- STEP 04
Give ChatGPT a sample of your own voice when possible, then ask it to remove robotic phrasing without changing the meaning.
/HUMAN Rewrite this caption in my natural, conversational voice. Preserve the message and remove robotic wording: [paste caption and an example of your writing]
- STEP 05
Let ChatGPT guide you with one question at a time instead of immediately handing you the answer.
SOCRATES Teach me [topic] by asking one question at a time. Use my answers to identify gaps and guide me toward the conclusion.
- STEP 06
Apply direct-response offer principles to make the outcome, proof, value, speed, and ease of an offer clearer.
HORMOZI Here is my offer: [paste offer]. Diagnose what makes it weak, then rebuild it so the value is clearer and the result feels more believable.
- STEP 07
Ask for the strongest opposing case, the missing evidence, and the failure modes that could break the idea.
RED TEAM Here is my plan: [paste plan]. Attack every major assumption, rank the failure modes, and show me the three best ways to make it stronger.
- STEP 08
Future You is a reflection exercise, while Lindy Mode filters advice through durable principles rather than whatever is currently trending.
FUTURE YOU LINDY MODE My goal is [goal]. From the perspective of my wiser future self, separate the durable principles from short-term noise and give me the next concrete move.
- STEP 09
Use ELI10 to make a difficult idea easy to understand, or YC Mode to interrogate a startup's problem, users, distribution, retention, and evidence.
YC MODE ELI10 Pressure-test this startup idea: [describe it]. Explain the biggest problem in plain language, then ask for the missing evidence you need.
- STEP 10
Combine two compatible codes when you want both perspectives in the same response.
80/20 RED TEAM Here is my launch plan: [paste plan]. Find the highest-leverage actions, attack the weak assumptions, and rebuild the plan around what is most likely to work.
- STEP 11
This second setup prompt adds Blind Spot, X10Think, KillCritic, and Alt3 alongside /Human, giving you more reusable ways to critique and vary an answer.
In this conversation, treat these five codes as my custom prompt shortcuts. They are not built-in ChatGPT commands. Apply the matching instruction when I put a code before or after a request: /HUMAN: Write or rewrite the answer in natural, specific, conversational language. Remove robotic wording, filler, and generic transitions without changing the meaning. BLIND SPOT: Point out what I have missed, the assumptions I am making, and where the idea could fail. Explain the most important risk and how to test it. X10THINK: Go beyond the first obvious answer. Examine the key trade-offs, propose a stronger approach, and explain why it is better. Do not just make the response longer. KILLCRITIC: Challenge my idea constructively instead of agreeing by default. Identify what is weak, what evidence is missing, and the clearest way to improve it. ALT3: Give me three meaningfully different versions of the answer. Label each version and explain its best use in one sentence. If I use more than one code, combine their instructions where possible. Ask only if they conflict or you lack essential information. Do not claim these codes are official ChatGPT features.
A FEW THINGS THAT HELP
Get more out of it.
- These are custom prompt cues, not built-in ChatGPT commands or guaranteed magic words. Define them first for more consistent results.
- If /HUMAN opens a slash-command picker in your interface, type HUMAN: as ordinary text instead, with the same definition.
- You do not need every code. Save the five to ten that match the work you repeat most often.
- Give ChatGPT real context after the code. Better inputs still produce better answers.
- For any code from the longer list, explain what you mean in plain language or define it in a setup prompt before relying on it.