🏴‍☠️ AI Is Dumb Until You Do This

The Missing Link To AI Mastery

💡 The One AI Skill That Separates You From Everyone Else

I was frustrated.

I’d spent hours trying to get ChatGPT to give me the answer I needed, but it kept failing.

It felt like owning a supercar but not knowing how to drive it.

Then, I stumbled upon a lecture by Andrej Karpathy that changed everything.

In partnership with

2x conversions by pre-testing your ads? Yes, it's possible!

  • Get instant AI feedback on your campaign creatives.

  • Improve brand visibility and memorability.

  • Tweak ads for maximum engagement and performance.

It wasn’t about AI theory; it was about how to actually make AI work.

There was no fluff, no jargon, just real, practical techniques.

You can watch his breakdown right here, but if you want the distilled, usable version, keep reading.

🔄 How AI Thinks (and Why It Matters)

AI doesn’t “understand” things like humans. Instead, it predicts patterns based on vast amounts of data.

Knowing this unlocks how to craft better prompts so you receive exactly what you need.

  • ✨ Base models are like parrots—they repeat patterns but lack real conversation skills.

  • 🌟 Instruct models (like ChatGPT) are trained with feedback to follow instructions better.

  • 📈 Your results depend on how well you design your prompts.

🔮 Mastering Prompts for Maximum Output

Think of prompting as giving directions to a genius—be clear, structured, and specific.

  • Few-shot prompting: Show examples before asking your question.

    "Q: Capital of France? A: Paris Q: Capital of Italy? A: [model completes]"

  • Role-based prompting: Have AI step into an expert’s shoes.

    "You're a personal trainer. Design a weekly workout plan for someone who wants to build muscle with minimal equipment."

  • Structured templates: Break down your task into clear, organized components.

    *"Create a meal plan for a week with:

    • 3 vegetarian dinners

    • 75 grocery budget

    • Under 30-minute prep time."*

Want to see if your prompt is too long?

Measure it with TikTokenizer to optimize its length as prompt up to 1,000 tokens work the best.

This isn’t traditional business news

Welcome to Morning Brew—the free newsletter designed to keep you in the know on the business news impacting your career, company, and life—in a way you didn’t know you needed.

Note: this isn’t traditional business news. Morning Brew’s approach cuts through the noise and bore of classic business media, opting for short writeups, witty jokes, and above all—presenting the facts.

Save time, actually enjoy business news, and join over 4 million professionals reading daily.

🎨 Creativity vs. Accuracy: Controlling Temperature

Not every AI task needs the same level of randomness. The temperature setting controls how predictable or creative the AI’s responses are:

  • 0.0 → For precise, factual answers (coding, research).

  • 0.3 - 0.7 → Balanced mix of logic and creativity (writing, brainstorming).

  • 0.8+ → High randomness, best for storytelling, idea generation.

If you need a reliable answer, set it lower. If you’re brainstorming fresh ideas, turn it up.

🔍 Real Uses (That Actually Help You)

📘 Personal Research Assistant

Prompt:

"Summarize the key differences between Lithium-ion batteries, Hydrogen fuel cells, and Pumped hydro storage. Include efficiency metrics and environmental impact."

💻 Debugging Code Like a Pro

Got an error? Instead of pasting vague questions, provide context.

def calculate_average(nums):
    total = 0
    for i in range(len(nums)):
        total += nums[i]
    return total / len(nums)

# Error: ZeroDivisionError when empty list passed

Adding the error message helps AI diagnose and fix it faster.

📑 Verifying Information (No More Fake Facts)

Prompt:

"List 3 peer-reviewed sources confirming that CRISPR-Cas9 can edit mitochondrial DNA. If none exist, say 'No verified sources found'."

This method helps prevent AI hallucinations—where it confidently gives wrong answers.

🔒 Preventing Misinformation (Fact-Prove Like a Pro)

If an AI response sounds questionable, run a fact-checking workflow like this:

graph LR
A[LLM Response] --> B{Critical Claim?}
B -->|Yes| C[Cross-evaluate with Google Scholar]
B -->|No| D[Use as-is]
C --> E[Update Response]

🔄 Final Tweaks for Optimal Results

Task

Best Model

Token Limit

Temperature

Research

Instruct Model

1,500

0.2

Storytelling

Base Model

4,000

0.9

Coding

Code Model

8,000

0.2

🔑 Unlocking Advanced AI Techniques

  • For coding help: Paste the exact error message and specify the language version.

  • For complex problems: Use chain-of-thought prompting:

    "Explain step by step before answering: [insert question]."

  • For repetitive work: Have AI generate templates.

    "Create a weekly planner with: Priorities, Deadlines, and Notes sections."

📈 Conclusion: The AI You Design is the AI You Use

AI isn’t magic—it’s a tool. And like any tool, how well you use it determines your results. Want to go deeper? Watch Karpathy’s full breakdown right here.

The AI revolution isn’t coming. It’s already happening.

The question is, how well are you using it?

Tool of the Day: Riffusion Fuzz

What It Does:
Riffusion Fuzz is an AI-powered music creation platform that lets users generate unique music based on their imagination.

Why It’s Awesome:
Currently in beta, it showcases its capabilities with four demo songs in different genres: modern hip-hop with funk elements, indie pop with atmospheric effects, melodic trap, and French house with techno influences.

Pro Tip:
Experiment with different prompts to discover unexpected musical styles and textures. Riffusion is continuously evolving, making it an exciting tool for musicians, AI enthusiasts, and sound designers alike.

Prompt of the Day:

"You are an eccentric, overly enthusiastic AI assistant with a flair for the dramatic. My task is to provide you with a completely mundane, everyday situation from my life—like making toast, waiting in line, or tying my shoes. Your task is to transform this event into an epic, suspenseful, or hilariously exaggerated tale, as if narrated by a medieval bard, a sci-fi AI overlord, or an over-the-top sports commentator. The result should be an unexpected, amusing, and wildly entertaining retelling of my boring event. Let’s turn the ordinary into the extraordinary!"

Example Input:
"I spilled coffee on my shirt this morning."

Example Output:
"In the grand theater of fate, upon the battlefield of the breakfast table, a most tragic duel unfolded. The noble warrior—your crisp, pristine shirt—stood unflinching, unaware of the dark, malevolent forces swirling in the treacherous depths of your coffee cup. With one fateful misstep, the abyss struck! Molten liquid surged forth, staining the white fabric like the sorrowful ink of destiny! A battle was lost, but the war of laundry yet remained…"

Paste this into your favorite assistant, and you might uncover a time-saving gem.

Challenge of the Day: The Absurdly Specific Compliment Quest

Pick a friend, colleague, or even a stranger (within reason) and craft the most bizarrely specific yet undeniably flattering compliment possible. Think beyond the usual, praise their “shockingly efficient stapler technique” or their “legendary ability to untangle headphones on the first try.”

Need help? Use an AI tool (like ChatGPT) to generate an over-the-top, Shakespearean-style ode to their greatness. Deliver the compliment with complete sincerity and note their reaction. Did they laugh? Look confused? Start questioning reality? Share your best (or weirdest) results, we might just feature them in the next issue!