GitHub Copilot After Two Years: The Honest Developer Review
There's a lot of hype around AI coding tools. There's also a lot of backlash. After two years of daily use across multiple languages and project types, I think I've got a nuanced enough view to cut through both.
What Copilot Actually Does Well
The best use case for Copilot isn't what people usually talk about. Everyone focuses on "it writes code for you" — and yes, it does that. But the more valuable capability, in my experience, is how it reduces the cognitive load of remembering syntax and boilerplate.
When you're working in a language you know reasonably well but don't use every day, Copilot fills in the gaps. You know what you want to write; Copilot handles the exact syntax. That's genuinely useful.
For repetitive patterns — writing similar functions across a codebase, generating test cases, filling in CRUD operations — Copilot is excellent. It sees the pattern from your first few functions and replicates it accurately.
The Real Limitations
Copilot confidently generates wrong code. This is the thing junior developers need to hear most: the suggestions look correct. The formatting is clean, the function names make sense, it looks professional. And sometimes it's subtly broken.
You cannot use Copilot as a replacement for understanding what you're writing. If you accept suggestions without reading them carefully, you will ship bugs. I've done it. Everyone who uses Copilot long enough has done it.
For complex business logic, Copilot is often unhelpful. It's good at patterns it's seen in training data. Your company's specific requirements, domain logic, and architectural decisions? It guesses, and the guesses are often wrong.
Language and Framework Differences
Copilot works significantly better in some languages than others. JavaScript, Python, TypeScript, and Go get excellent suggestions. Less popular languages get noticeably worse output.
For React and Next.js specifically, it's quite strong — which makes sense given the enormous volume of React code it was trained on.
Is It Worth $10/Month?
For most professional developers: yes. The time saved on boilerplate and syntax lookup probably exceeds the cost within a week.
For beginners: be careful. The risk of learning to rely on suggestions without understanding them is real. Better to struggle through the understanding first.
For occasional coders: maybe not. The benefit comes from daily use where it learns your patterns.
My Setup
I use Copilot in VS Code with autocomplete enabled but I've turned off the "ghost text" feature that shows inline suggestions — I found it distracting. Instead, I use it on demand with keyboard shortcuts when I want a suggestion. This gives me the benefits without the constant visual noise.
Bottom line: It's a legitimate productivity tool when used by experienced developers who read what it generates. It's a crutch that creates problems when treated as an automatic code generator.