Using GitHub Copilot in VS Code: Friend or Friction?

👋 Introduction

Like many developers today, I often code in VS Code with the help of GitHub Copilot. It’s fast, it’s clever, and sometimes it’s... too clever.
I decided to reflect on my real-world use of AI in my daily workflow: what works, what doesn’t, and what still needs a human touch.

---

🧩 Where Copilot Truly Shines: Documentation

If there’s one task where Copilot really excels, it’s documentation updates.

• It reads the function signature and context.

• It generates relevant PHPDoc blocks, Markdown explanations, or even API docs with little effort.

• When I refactor a method, it automatically proposes accurate comments that evolve with the code.

🟢 Conclusion: For keeping documentation consistent and up to date, AI is a brilliant assistant — as long as I double-check the results.

---

🛑 Code Reviews: When AI Becomes Too Opinionated

I also tested AI-powered code review tools (like Copilot Suggestions or other extensions).
Honestly? That’s where the magic started to fall apart.

• The AI flags nearly everything: naming, spacing, architecture, structure, even when the code is clean.

• It blocks merge requests on stylistic details or differences in philosophy.

• It often ignores project-specific conventions in favor of some abstract “perfect code.”

❗ My take: Code reviews should help, not hinder. An AI should align itself with the project’s standards — not impose new ones.

---

🧪 New Features: Copilot is Only as Good as Your Prompts

When building new features, Copilot can save time — or waste it.

• If I guide it with clear prompts, it proposes good boilerplate code and clean structures.

• If I let it “improvise,” I get inconsistent logic, mismatched naming, and even code that breaks project conventions.

🧭 Lesson learned: Don’t expect the AI to follow your architectural style unless you teach it. Show it. Guide it. Prompt it.

---

🔍 The Prompt Matters More Than the Code

In all cases, there’s one recurring truth: the prompt is everything.

• Be precise. Example: create a Laravel service to send notification emails, using the existing EmailNotification interface and respecting PSR naming conventions.

• Give context. Don’t assume the AI knows your project.

• Always review, refactor, and question what Copilot generates.

---

💬 Final Thoughts

AI is a powerful ally — but not a replacement for your brain.

• Use it where it adds value: documentation, repetitive code, boilerplate.

• Be skeptical during code reviews or feature design.

• Always prompt intelligently and trust your human intuition.

“Copilot is your assistant, not your architect.”

---

📌 TL;DR

🎯 Tips for Prompting Copilot Efficiently

🧠 Think like a teacher, not a typist. Give Copilot clear instructions and context.

✅ Be specific

// Bad
"create a notification"

// Good
"Create a Laravel Notification class named OrderShippedNotification that uses MailChannel"

✅ Define structure

"Generate a DTO class with public readonly properties and a fromModel() method"

✅ Set context

"We use snake_case for DB columns and camelCase for variables. Don't use helper() functions; use services."

✅ Review & Refactor

• Never copy blindly

• Refactor the result to match your project

• Add your own business logic

🧩 AI needs clarity — the more context you give, the better the result.