5.2. GitHub Copilot: AI-Powered Code Suggestions
Copilot has grown from inline autocomplete into something closer to a collaborator, and the exam expects you to know the shape of what it can and can't do on its own.
💡 First Principle: GitHub Copilot generates AI-powered code suggestions grounded in the context of your project — comments, existing code, and open files — rather than producing generic snippets in isolation. Its most autonomous mode, Agent Mode, goes further: instead of suggesting one snippet at a time, it can break a task into subtasks and carry out multi-file changes on its own, proposing terminal commands and edits across a codebase. Copilot agents extend this further into more independent, decomposed problem-solving. Underneath all of this, multi-model support means Copilot isn't locked to one AI model — teams can choose from multiple providers depending on the task, rather than accepting a single one-size-fits-all model.
⚠️ Exam Trap: Copilot's suggestions — including Agent Mode's multi-file changes — are AI-generated starting points, not verified solutions. They still require human review and testing through the normal pull request process; more autonomy doesn't mean less oversight.
Reflection Question: A developer uses Agent Mode to implement a feature across four files. Does merging that change still require the same code review the GitHub Flow expects for any other pull request?
