Ghost complete while you type — great for daily functions. Repo-wide rewrites want a chat that can see the tree.
Copilot earns its keep while the caret moves: half a test name becomes assertions in the file’s style; half a regex finishes like the line above. That saved typing adds up. It will not choose an architecture. I disable “rewrite the file” impulses and keep inline and short blocks. Versus Cursor it is weaker at “spread this pattern across eight packages.” Versus Codeium it sits deeper in GitHub, PRs, and workplace procurement — and usually costs more. Treat completion as cloud unless policy says otherwise. Completion is an IME; review is mine. When it emits a stale API, the merger owns it. Clear job, good value. As lead engineer it produces plausible duplication.
People writing product code in VS Code or JetBrains every day. Full-time cross-repo refactors need a heavier chat.
Stays in the keystroke
Faster than pasting a file into chat when the style is already on screen.
Tests and boilerplate
A descriptive test name often buys a draft you can edit.
Lives in the IDE you have
Fewer editor wars.
Enterprise story
Easier to explain next to GitHub identities and permissions.
Short horizon
Cross-package moves come back locally right and globally duplicated.
Stale APIs
The index is not yesterday’s release. Check the docs.
Pricey for side projects
Personal work often has a cheaper complete engine.
Write the name and given/when/then; accept; fix asserts.
DTO soup gets a base; you fix nulls and names.
Two examples in, finish the pattern, test on real strings.
They change. Read GitHub today, not a screenshot.
It replaces some typing, not design fights or ownership.
Possible, noisy. I run one inline engine.
Depends on plan and contract. Companies should buy, not piggyback a personal seat.
Boilerplate-heavy languages feel better. Rare DSLs drift.
The link opens the GitHub Copilot site so you can check our judgment. The review is written by AI WebLink.