By: Dr. Hamza Mousa
What GenAI Really Needs? GenAI is Not a Magic Wand: It’s a Discipline, Do Not Blame AI For Your Bad AI Results!
Why Did I Write this Post?
I’ve seen many people use GenAI, but the results always confuse me. It’s as if they believe the tool has access to their mind, or can somehow understand the mental context they’ve built, without ever being told what they mean.
I’ve watched developers generate flawed codebases because they didn’t know how to guide the AI.
I’ve seen designers using Photoshop (or other creative tools) revert back to “no-AI” mode, not because AI is bad, but because they never learned how to work with it.
And I’ve seen business teams stuck doing the same manual work every day, not because automation isn’t possible, but because no one taught them how to write effective prompts or refine the output.
But the worst part?
Too many people, with little to no foundational knowledge, sense, or education, believe GenAI can turn them into scientists, artists, engineers… even doctors, in minutes.
That’s dangerous. And frankly, shameful.
GenAI tools aren’t magic. They’re a discipline, just like any other skill: coding, painting, surgery, horseback riding. They demand knowledge, education, consistent practice, real-world experience, and yes, a healthy dose of creativity and wild imagination.
So here, I’m writing about why you should invest in learning AI properly, and what you actually should do to get better.
What To Do, for Better Results? Learn the Discipline of GenAI!
1- Be Creative (Go Wild)
You don’t need permission to be creative. You don’t need a degree. You don’t need to wait for an AI to “inspire” you. The moment you start imagining something, a design, a story, a system, that’s where it begins.
But here’s the catch: AI doesn’t read your mind. It only responds to what you say.
So if you want magic, you have to build the spell first.
Don’t ask for “a beautiful logo.” Ask for “a minimalist logo for a mental health app, inspired by calm water and open skies, using soft gradients and a single line symbol.”
Be specific. Be visual. Be bold. Let your imagination run, then guide the AI with precision. Creativity isn’t about letting go. It’s about knowing exactly what you want, and demanding it from the tool.
2- Be Precise, yet Speak Context
Vagueness kills results, check the following prompt:
- “Make it good” = nothing.
- “Fix this code” = chaos.
- “Design a website” = endless confusion.
Precision is power, it is also the base of every discipline.
Tell the AI what kind of result you want, not just that you want it.
Use structure. Use constraints. Use details.
Instead of “write a report,” say: “Write a 500-word technical report on AI in drug discovery, focused on RAG systems, with citations from 2022–2024, in plain language for non-experts.”
The more precise you are, the less guessing the AI has to do.
And the fewer hours you waste fixing garbage output. Precision isn’t boring. It’s the foundation of control. When you’re clear, the AI becomes your co-pilot, not your boss.
3- Understand then Plan your End Result!
Before you type a single prompt, ask: What does success look like?
Not “something useful.” Not “a draft.”
But exactly what will make you say, “Yes, this is it.” A final product? A prototype? A pitch deck? A debugged script?
Define it. Visualize it. Break it down.
Then reverse-engineer the steps. If you want a database schema, don’t ask for “a table.” Ask for “a PostgreSQL schema with users, roles, permissions, timestamps, and foreign keys linking to projects.”
Plan the outcome ahead, before the input. Because AI doesn’t plan for you. It executes. And if you don’t know where you’re going, it’ll take you anywhere.
Clarity comes before creation.
4- Know what you are doing! (Study)
You don’t need to be an expert, but you do need to know what you’re asking for. If you’re generating code, know what a function does. If you’re designing a UI, know what makes a layout usable. If you’re analyzing data, know what a correlation means.
AI won’t tell you when you’re wrong. It’ll just give you a polished lie.
Understanding the basics, even at a surface level, turns you from a passive user into an active thinker.
You’re not just feeding inputs. You’re testing ideas. Validating logic. Refining outcomes.
When you know what you’re doing, you spot errors fast. You reject bad output. You improve the next try. Knowledge isn’t optional. It’s your shield against hallucination.
5- Know your Terms and Structure, (Your AI Language)
AI doesn’t speak human, it speaks structured language.
So learn the terms: prompt, token, model, fine-tune, RAG, embedding, inference, context window, temperature, top-k. Not to impress anyone. To understand what the tool is doing. When you know what “temperature” means, you can adjust it to get more creativity or more consistency.
When you know what a “context window” is, you won’t flood the AI with 100 pages of text. Knowing terms isn’t academic, it’s practical.
It’s how you avoid wasting time, misusing features, or trusting nonsense.
Start small. Learn one term a week. Build your vocabulary.
The more you know, the more you control.
6- Visualize in your head
Before you type anything, close your eyes.
See the result. Imagine the color scheme. The layout. The tone. The flow.
The shape of the file. The sound of the voice. The rhythm of the sentence.
This mental image is your compass.
It’s what guides your prompts. If you can’t picture it, the AI can’t deliver it.
Visualization isn’t fantasy, it’s focus.
Build with purpose.
It forces you to clarify your goal, it separates dreams from reality. And when you bring that vision to the prompt, the AI doesn’t guess, it follows.
Your mind is the blueprint. The AI is the builder.
7- Variations (Always Ask for More and Compare)
One prompt ≠ one perfect result. Always ask for variations:
- “Give me 3 different versions of this email.”
- “Show me 2 alternative layouts for this dashboard.”
- “Generate 4 headlines, one serious, one playful, one urgent, one poetic.”
Variation isn’t waste. It’s exploration, it gives you options. It reveals patterns. It helps you see what works, and what doesn’t.
Don’t settle on the first output. Test. Compare. Choose.
Even if you pick one, the others teach you something.
AI is not a single answer machine. It’s a brainstorming partner.
Use it to stretch your thinking, not to outsource your judgment.
8- What IF!
Ask “What if?”:
- What if we tried this structure?
- What if the tone was sarcastic instead of formal?
- What if we removed the intro and started with the conclusion?
- What if the AI used metaphors instead of facts?
“What If” is the spark of innovation, it breaks loops. It challenges assumptions.
It pushes the AI beyond default responses. Don’t fear weird ideas. Try them. You might find a gem. Or at least learn what doesn’t work. Curiosity is free. Experimentation is powerful. Let your imagination test the limits, then refine what survives.
9- Why AB Testing is important!
You wouldn’t launch a product without testing.
So why trust an AI output without comparing it? AB testing isn’t for marketers only, it’s for every creator.
Run two versions of the same prompt.
Compare the results:
- Which one’s clearer?
- More accurate? More engaging?
- Which one solves the problem faster?
Then, keep the winner, Improve the loser. This isn’t overkill, it’s discipline.
It turns random output into reliable insight, it builds confidence in your process. And it turns GenAI from a guess into a tool you can trust.
Test. Learn. Repeat. That’s how mastery grows.
10- Know your tools, variations, strength, weaknesses
Every AI tool has its own rhythm, its own personality. Ollama runs locally, a quiet guardian of your data, perfect for private research, though it slows down when handling massive models. OpenAI? Lightning-fast, but tethered to the internet, always listening.
Anthropic delivers thoughtful, nuanced responses, yet comes at a premium price. LM Studio offers flexibility and deep control, but setting it up can feel like untangling a complex knot. No single tool is universal. They each speak a different dialect.
So don’t treat them all as equals. Know what you’re asking for before you ask. Test them side by side. Compare outputs. Learn their quirks, the way one stumbles on technical details, another overcomplicates simple tasks. Use Ollama when privacy matters most. Turn to OpenAI when speed is king.
Let LM Studio shine when you need precision and local autonomy. Each has strengths, yes, but also blind spots. The real skill isn’t in picking the “best” model, but in knowing which one fits this task, this moment.
Final Thought
GenAI doesn’t replace you, it reveals you. The better you are at thinking, the better it becomes at helping. Skill isn’t in the tool. It’s in the mind behind it. So stop waiting for magic, start building clarity. Because mastery isn’t about using AI, it’s about leading it.
