GitHub Copilot is the most widely used AI coding assistant and the tool that started the AI pair-programming conversation. After a year of sustained use across multiple projects, here is what I have found.
What Copilot Does Well
Autocomplete at the function level — suggesting entire functions based on the function name, existing code context, and comments — is Copilot’s core value. It is particularly strong for: writing boilerplate code (getters/setters, serialization), completing patterns you’ve already established (once you write three similar functions, it predicts the fourth), test generation from existing functions, and translating pseudocode comments into working code.
Copilot Chat
The inline chat feature (available in VS Code and JetBrains IDEs) lets you highlight code and ask questions, request refactors, or have it explain what the code does. This is more useful for learning or for reviewing unfamiliar code than for primary development work.
Where It Falls Short
Copilot has limited awareness of your full codebase — it works best with the currently open file and nearby files. For suggestions that need to respect complex architectural decisions across many files, it often produces plausible-but-wrong code. Security-sensitive code generation is also risky — Copilot has been found to suggest vulnerable patterns for authentication and cryptography.
Alternatives
Cursor (VS Code-based) and Zed (new editor with native AI) have surpassed basic Copilot for many developers due to better codebase awareness. Copilot retains advantages in IDE integration breadth and GitHub ecosystem hooks.
Pricing
Copilot Individual is $10/month ($19/month for Pro with more models). Free for verified students via GitHub Education. Worth the cost if you write code daily.
