Use Copilot for a tightly managed coding day

A commenter had GitHub Copilot produce all the code they pushed in one day, with close supervision throughout.

Done with GitHub Copilot· Intermediate

What they did
A commenter said they let GitHub Copilot generate every line they committed that day, but only while they kept directing it step by step. In the same discussion, other people described a careful routine: give the agent a very narrow goal, say what must stay unchanged, check the code before accepting it, and fall back to backups if it starts breaking things.
How it went
The setup was good enough for that day’s work, but it depended on constant human oversight. The thread’s wider conclusion was that agents can struggle on larger tasks and may force a full revert or restart when they go off course; one person said a game change needed 2 days of trial, error, and rollbacks before it finally worked.
Tips
Keep each request small, and make the no-touch areas explicit. Review and test every step, and have a backup ready in case the agent edits the wrong parts.

Try it yourself with GitHub Copilot

Help me use GitHub Copilot on [repo/project] to implement [feature or fix] in small pieces. For each piece, make the smallest safe edit, then report the files changed and what you altered so I can review it before you continue. Leave [areas to avoid] alone, and if a change would affect existing behavior, flag that clearly before proceeding. Pause for my approval before any irreversible step such as applying, submitting, deleting, or sending anything. Stop when [done condition] is met and the relevant tests pass.

Fill in the [brackets]. Agents can make mistakes, so review anything before it is sent, paid or deleted.

Read the original on reddit.com ↗

Get 3 agent uses like this every morning

With the day's top AI news. Free, and one click to unsubscribe.