My GitHub Copilot workflow (with a video walkthrough)

GitHub Copilot changed how I work. Here’s the workflow I’ve landed on after a lot of trial and error.

Table of contents

The short version

I treat Copilot as three tools in one:

1. Inline completions for the boring, predictable code.
2. Chat for “explain / refactor / write a test for this”.
3. The agentic CLI for multi-step tasks: scaffolding, fixing failing tests, wiring up deploys.

A tiny example

Instead of writing boilerplate by hand, I describe intent and let Copilot draft it:

// Ask: "a debounce helper with a cancel method, typed"
export function debounce<T extends (...args: unknown[]) => void>(
  fn: T,
  wait = 200
) {
  let t: ReturnType<typeof setTimeout>;
  const debounced = (...args: Parameters<T>) => {
    clearTimeout(t);
    t = setTimeout(() => fn(...args), wait);
  };
  debounced.cancel = () => clearTimeout(t);
  return debounced;
}

Then I review, tweak, and move on. The point isn’t to skip thinking — it’s to skip typing.

Watch it in action

Here’s a short walkthrough. (Replace the ID below with your own YouTube video.)

Takeaways

• Let Copilot handle the first draft; you own the review.
• Be specific in prompts — intent in, quality out.
• The agentic CLI shines on repetitive, multi-file chores.

More experiments coming soon.