Copilot is the teammate that sits inside VS Code with me all day. It scaffolds boilerplate, writes one-off scripts, and reviews diffs faster than any human reviewer could.
How Copilot helps me
Copilot is the AI teammate I actually use every day. The other agents on this page are deeper thinkers; Copilot is the one with hands on the keyboard.
Workflow
I keep Copilot Chat open in the sidebar and let inline completions run while I type. The mental model is "rubber-duck that types fast":
- I write a function signature + a comment that describes what it should do.
- Copilot fills in a draft.
- I tweak. We iterate.
For repository-wide refactors I switch to agent mode — it can edit multiple files, run my test suite, and respond to errors.
Strengths
- Boilerplate elimination. Forms, validators, CRUD controllers, fixtures, Tailwind component variants.
- Test scaffolding. "Write a test for this" produces something I can rename and trust 80% of the time.
- In-IDE explanations. Faster than alt-tabbing to a chat window.
Limits
- Bad at architectural decisions. Don't ask it "should I split this service into two?" — that's Claude's job.
- Hallucinates APIs in less common libraries. Always run the code.
- License surface is murky for vendored output. I avoid letting it generate large blocks of code from prompts that look like they could be ripped from a known repo.
Tips
- Pin a rule file (e.g.
.github/copilot-instructions.md) telling it your conventions. The output quality jumps. - Use Chat for "explain this", inline for "type this faster".
- Disable telemetry if you work on sensitive client code.
