A reviewer asks how a Copilot task can move from a request to a proposed patch. Which description best captures the distinction between generation and agentic execution?

Published
  1. Every generated code block is automatically executed in the production application.
  2. Tool permissions become unnecessary as soon as a language model is selected.
  3. The model can propose content, while the surrounding agent uses permitted tools to inspect, edit, and validate the project.
  4. Running tests updates the model weights to guarantee that later patches pass.

Comments — 0