The most powerful AI right now are the reasoning models.
But only 0.01% know how to unlock the full potential of o1, o3, Gemini 2.5, and Claude Sonnet 3.7.
That's why we created this cheatsheet to help you.
Here's everything you need to master prompt writing:
But only 0.01% know how to unlock the full potential of o1, o3, Gemini 2.5, and Claude Sonnet 3.7.
That's why we created this cheatsheet to help you.
Here's everything you need to master prompt writing:
First, assign the model a thinking role.
This changes how it “shows up” mentally.
Examples:
- Strategy Consultant
- Logical Analyst
- Detective
- Policy Evaluator
- Research Synthesizer
The more specific the role, the sharper the output.
This changes how it “shows up” mentally.
Examples:
- Strategy Consultant
- Logical Analyst
- Detective
- Policy Evaluator
- Research Synthesizer
The more specific the role, the sharper the output.
Next, give it a reasoning task.
This tells the model what kind of thinking to perform.
Try these:
- Tradeoff Analysis
- Root Cause Breakdown
- Ethical Scenario Assessment
- Heuristic Decision Tree
- Socratic Dialogue
Be explicit.
This tells the model what kind of thinking to perform.
Try these:
- Tradeoff Analysis
- Root Cause Breakdown
- Ethical Scenario Assessment
- Heuristic Decision Tree
- Socratic Dialogue
Be explicit.
Then set mental constraints to make it think rigorously.
This is what separates “smart” from “strategic.”
Use constraints like:
- Step-by-step logic
- First principles
- Compare 3 POVs before concluding
- Show your reasoning trail
Avoid emotional language
This is what separates “smart” from “strategic.”
Use constraints like:
- Step-by-step logic
- First principles
- Compare 3 POVs before concluding
- Show your reasoning trail
Avoid emotional language
Choose your output format.
Why? Because format shapes thinking.
Ask for:
- Flowchart
- Logic trail
- Table of pros/cons
- Decision matrix
- Compare/contrast chart
- Hypothesis tree
The model adapts to structure.
Why? Because format shapes thinking.
Ask for:
- Flowchart
- Logic trail
- Table of pros/cons
- Decision matrix
- Compare/contrast chart
- Hypothesis tree
The model adapts to structure.
Now you’re ready to write the actual prompt.
Just use the R.E.A.S.O.N. formula:
R – Role
E – Evaluate
A – Ask Questions
S – Simulate
O – Output format
N – Narrow the scope
Every powerful prompt is a remix of these.
Just use the R.E.A.S.O.N. formula:
R – Role
E – Evaluate
A – Ask Questions
S – Simulate
O – Output format
N – Narrow the scope
Every powerful prompt is a remix of these.
Here are a few plug-and-play prompt templates to try:
“Act as a logical analyst. Evaluate 3 options for [X]. Use only deductive reasoning and output a step-by-step decision matrix.”
“Compare 3 strategies to reach [goal]. Rank them by ROI, and explain using first principles.”
“Act as a logical analyst. Evaluate 3 options for [X]. Use only deductive reasoning and output a step-by-step decision matrix.”
“Compare 3 strategies to reach [goal]. Rank them by ROI, and explain using first principles.”
This cheatsheet works with all major reasoning models:
- GPT-4o
- Claude Sonnet 3.5 / 3.7
- Gemini 1.5 / 2.5
- Mistral
- Meta's LLaMA
It’s not about the model.
It’s about the mental structure you give it.
- GPT-4o
- Claude Sonnet 3.5 / 3.7
- Gemini 1.5 / 2.5
- Mistral
- Meta's LLaMA
It’s not about the model.
It’s about the mental structure you give it.
I hope you've found this thread helpful.
Follow me @jonsant0s for more.
Like/Repost the quote below if you can:
Follow me @jonsant0s for more.
Like/Repost the quote below if you can:
https://twitter.com/1783144394758946816/status/1937785307346407808
• • •
Missing some Tweet in this thread? You can try to
force a refresh
