1.1.1. From Autocomplete to Agentic Pair Programmer
💡 First Principle: Copilot's capabilities form a ladder of autonomy. At the bottom, it completes the line you're typing. At the top, it plans a task, edits many files, runs commands, and checks its own work — with you approving the plan and the result.
When Copilot first shipped, the whole product was inline suggestion: ghost text completing the current line or block. That capability still exists and is still the most-used feature, but the January 2026 exam rewrite exists precisely because the product grew past it. Understanding the ladder helps you place any feature the exam mentions:
| Rung | What it does | Human role |
|---|---|---|
| Inline suggestions | Completes current line/block as you type | Accept, reject, or cycle alternatives |
| Copilot Chat | Answers questions, explains code, drafts snippets in a chat panel | Ask, refine, copy in |
| Edit Mode | Applies scoped edits across multiple files you specify | Direct each turn, review diffs |
| Plan Mode | Drafts a step-by-step plan before acting | Approve or adjust the plan |
| Agent Mode | Autonomously plans, edits, runs tools/terminal, and iterates | Set the goal, approve actions/results |
Consider a real scenario: a developer asks Copilot to "add input validation to all the API handlers and write tests." Inline completion cannot do this — it only sees one cursor position. Edit Mode can apply the changes if the developer points at the files. Agent Mode can discover the handlers, edit them, generate tests, and run them, pausing for approval. The exam likes to test whether you can match the right rung to the task.
⚠️ Exam Trap: Don't equate "newer" with "always better." Inline completion is still the correct, lowest-overhead choice for finishing a known line of code. Agent Mode for a one-line completion is overkill and burns more resources. The exam rewards choosing the lightest capability that solves the problem.
Reflection Question: A teammate says, "Copilot can't really change more than one file at a time." Which capabilities would you point to, and what makes them different from inline suggestions?
