DEV Community

The Accessible AI Hub
The Accessible AI Hub

Posted on

Prompt Like a Pro, Pair Like a Legend

Why GitHub Copilot Might Just Be Your New Pair Programmer

Let’s be real. Talking to AI felt like a scene out of sci-fi movies. Now? It's just Wednesday. Welcome to the golden age where typing a fancy comment can summon a helpful (and slightly overenthusiastic) AI coder from the cloud.

First — What Even Is Prompt Engineering?

Imagine this: You walk into a café and ask for “coffee.” You get a cold brew with five pumps of vanilla. Not what you meant? That’s on you, buddy.

Prompt engineering is basically giving AI a clear, non-vague instruction. It’s not coding — it’s talking nicely to your AI friend so it doesn’t give you something weird like “temp func” instead of “Convert Celsius to Fahrenheit.”

Good prompt:
# Build a weather app that asks for a city, uses OpenWeatherMap API, and displays Celsius

Bad prompt:
# idk just do the thing

Think of prompt engineering as the art of giving instructions to your very literal junior dev (aka Copilot). Get this right, and you’ll feel like an AI whisperer. Get it wrong, and... well, enjoy debugging what you didn’t ask for.

Prompt Engineering: Your New Developer Superpower (That No One Taught You in College)

Let’s be honest — writing prompts feels like casting spells. Except instead of Latin incantations, you’re whispering sweet, structured nothings into the ears of a machine learning model.

And the scary part? How you phrase your prompt can make or break your entire solution.

Prompting isn’t just about typing # create function. That’s like texting your crush “hi” and expecting them to write you a novel. Nah — we’re past that era.

So... What Makes a Good Prompt?

Imagine you’re training a very smart but incredibly literal intern.

🧠 You: “Make a dashboard.”
🧑‍💻 Intern: “Here’s a blank HTML page and a broken button.”

Now try this:

“Build a responsive dashboard with cards showing monthly revenue, user count, and bounce rate. Use Bootstrap. Include dark mode toggle.”

Boom. That’s what we mean by prompt precision.

Prompting Tips That Actually Work:

💡 1. Use Comments as Commands The best way to prompt Copilot is by writing a comment like you’re describing what you need out loud.

# Write a function to sort a list of dictionaries by 'age' key in descending order
Enter fullscreen mode Exit fullscreen mode

💡 2. Be Specific, Not Cryptic Avoid vague terms like “do stuff” or “check this.”

// Validate email input using regex and show error message if invalid
Enter fullscreen mode Exit fullscreen mode

💡 3. Use Step-by-Step Thinking If the problem is complex, break it down. Copilot loves clarity.

# Step 1: Get user input
# Step 2: Fetch data from API
# Step 3: Parse response and display result
Enter fullscreen mode Exit fullscreen mode

💡 4. Add Partial Code Sometimes, showing a little code helps Copilot follow your thought process.

function calculateTax(income) {
  const taxRate = 0.2;
  // Now calculate the final tax based on deductions
}
Enter fullscreen mode Exit fullscreen mode

💡 5. Name Your Variables Wisely If your variables are named x, y, and temp2, Copilot might freak out (rightfully). Try:

# Calculate final price after applying discount
original_price = ...
discount_percentage = ...
Enter fullscreen mode Exit fullscreen mode

Prompting Is Not Just for Copilot — It’s the Language of All LLMs

Prompting isn’t a one-trick pony. Whether you’re using:

  • GitHub Copilot
  • ChatGPT
  • Google Gemini
  • Or even building with OpenAI APIs

Prompting is how you talk to the machine. Do it right, and AI turns into a rockstar teammate. Do it wrong, and you’re stuck in a loop of, “That’s not what I meant, bro.”

Now Enter: Pair Programming (With Less Awkward Eye Contact)

Old-school paired programming? You had two devs:

  • One typing (Driver)
  • One thinking (Navigator)

It was productive, enlightening, and just a tiny bit exhausting — especially when one person thinks in tabs and the other in spaces.

But here's the twist: you can now replace that second human with GitHub Copilot. No coffee breaks. No passive-aggressive sighs. Just code suggestions on tap.

🧠 You: Driver
🤖 Copilot: Navigator

You steer. It helps you get there faster (without judging your typos).

GitHub Copilot – Your AI-Paired Programmer

GitHub Copilot (made by GitHub + OpenAI) is not just autocomplete on steroids. It’s more like an AI intern with superhuman memory and zero salary expectations.

Inside your favorite editor, it reads what you’re typing and suggests:

  • Full functions
  • Boilerplate code
  • Tests
  • Dad jokes (okay not yet, but give it time)

Just type something like:

# Function to check if a number is prime
Enter fullscreen mode Exit fullscreen mode

...and BAM — Copilot fills in the blanks like a mind reader that majored in CS.

Don’t Worship the Bot — Know Its Limits

Even AI-powered sidekick has its quirks:

🟢 Strengths:

  • Boilerplate automation
  • Refactoring like a champ
  • Exploring unfamiliar APIs
  • Writing unit tests (you know you weren't going to)

🔴 Weaknesses:

  • Doesn’t get your “big picture” business goals
  • Sometimes hallucinates functions that sound cool but don’t exist
  • Doesn't remember past chats (yet — sorry, no continuity)

You’re still the developer. Copilot just makes you faster.

Final Word: Prompting Is the New Programming

Prompting isn’t just how you get AI to code — it’s how you communicate with the future. Copilot is here to help, but the real magic? It still comes from you knowing what to build and why.

So go ahead — type that smart comment, fire up Copilot, and build like a tech wizard with a robot sidekick.

🧑‍💻 You: Developer🤖 Copilot: Overachieving intern🔥 Result: Productivity x100

Until then, prompt wisely, pair proudly, and code like no one’s judging your semicolons.

Written by

Leerish Arvind
Beta Microsoft Student Ambassador | Co-Founder @ The Accessible AI Hub

© 2025 The Accessible AI Hub. All rights reserved.

Top comments (0)