How to rock: The Ultimate Guide to Creating Better AI Prompts
2025-12-03[ ai how-to-rock ]
After a couple of nights fighting with LLMs, I noticed that most "bad outputs" were actually my lack of clarity. This post is a brain dump so you can avoid my detours and ship prompts that behave.
TLDR: personas + context + structure beat "hey, write this for me" every single time.
The core principle
AI prompting is programming with words, not asking questions. You're not chatting, you are instructing a probabilistic completion engine. Treat prompts like mini specs: inputs, constraints, outputs.
If you can't explain what you want to yourself, the model will improvise (is this a feature or a bug ? ¯\_(ツ)_/¯). All prompting problems are thinking problems. Write the gist in your notebook, then paste it.
Essential techniques
- Use personas (ref)
- Tell the AI who it is and who it's talking to. This helps cut the noise.
- Provide context (ABC: Always Be Contexting) (ref)
- Include all relevant facts, even the obvious ones.
- Give explicit permission for the AI to say "I don't know", this reduces hallucinations.
- Remember: any missing context, the AI will fill in with guesses.
- Define output requirements
- Specify the format, tone, and any other constraints for the response.
- Use few-shot examples
- Show 2–3 written samples in the style or format you want. The more concrete, the better.
Advanced techniques
- Chain of Thought (COT) (ref)
- Ask the model to show its reasoning steps before writing the final answer.
- Tree of Thought (TOT) (ref)
- Request multiple solution options, then ask for a synthesis or best-of.
- Battle of the bots (adversarial validation) (ref:D3, ref)
- Set up competing personas and add a "critic" to help validate or improve outputs.
References
What should you do now ?
Maybe using AI to create the prompt for you is the best thing to do :)
For that, use the following text as an instruction for llms to write the prompt for you:
(Think of it as a blog inside of a blog)
# The Ultimate Guide to Creating Better AI Prompts
## The Core Principle
**AI prompting is programming with words, not asking questions.** You're not having a conversation—you're instructing a prediction engine that completes patterns based on probability.
## The Meta-Skill: Clarity of Thought
Before you write a single prompt, **think clearly about what you want**. If you can't explain it clearly to yourself, you can't prompt it. All prompting problems are actually thinking problems.
---
## Essential Techniques
### 1. **Use Personas**
Tell the AI who it should be to narrow its focus and improve responses.
**Example:** "You're a senior site reliability engineer for CloudFlare. You're writing to both customers and engineers."
**Why it works:** Gives the AI a specific perspective and expertise to draw from instead of generic knowledge.
---
### 2. **Provide Context (Most Important)**
Give the AI ALL the details it needs. Never assume it knows something—more context = fewer hallucinations.
**Key principle:** ABC = Always Be Contexting
- Include all relevant facts
- Be specific and detailed
- Don't hold back information
- Give it permission to say "I don't know" if it lacks information
**Pro tip:** Whatever context you don't provide, the AI will fill in with guesses.
---
### 3. **Define Output Requirements**
Tell the AI exactly how you want the result to look.
**Specify:**
- Format (bulleted list, under 200 words, etc.)
- Tone (professional, casual, anxious, etc.)
- Structure (timeline, sections, headers)
- Style constraints (no corporate fluff, radically transparent)
---
### 4. **Use Few-Shot Examples**
Show the AI what you want instead of just describing it. Provide 2-3 examples of the exact output style you're looking for.
**Why it works:** Removes guesswork—you're teaching by example rather than description.
---
## Advanced Techniques
### **Chain of Thought (COT)**
Ask the AI to think step-by-step before answering.
**Example:** "Before writing this email, think through it step by step..."
**Modern shortcut:** Use "Extended Thinking" or "Reasoning" mode built into most AI platforms.
---
### **Tree of Thought (TOT)**
Have the AI explore multiple approaches simultaneously, then synthesize the best path.
**Example:** "Brainstorm three distinct approaches: [A], [B], [C]. Evaluate each, then synthesize them."
---
### **Battle of the Bots (Adversarial Validation)**
Create competing personas that critique each other.
**Example:** "Generate a 3-round competition with [persona 1], [persona 2], and [critic]. Have them debate and collaborate on the final output."
**Why it works:** AI is better at critiquing than original creation.
---
## Common Mistakes to Avoid
1. **Being vague** - AI fills gaps with guesses
2. **Skipping context** - Leads to hallucinations
3. **Assuming knowledge** - AI doesn't remember everything
4. **Not defining output** - Gets generic results
5. **Messy thinking** - Creates messy prompts
---
## The Process
1. **Think first, prompt second**
2. Write down exactly what you want before prompting
3. Ask yourself: "Could a human do this with the information I've provided?"
4. If yes → the AI probably can too
5. If no → add more context and clarity
---
## Quick Wins
- **Enable web search** when you need current information
- **Save successful prompts** in a library for reuse
- **Use prompt enhancers** to improve your raw prompts
- **Treat bad outputs as skill issues** - the problem is usually your clarity, not the AI
---
## Remember
The AI can only be as clear as you are. When frustrated with results, don't blame the AI—look in the mirror. It's a skill issue. Get clearer about what you want, and the AI will get clearer in its responses.