We’ve all been there — Rebel starts doing something unexpected, and our instinct is to type “No, stop — I actually meant…” and correct course in the same conversation.
It works, but there’s a better way: edit your original message and rerun it.
Why does this matter?
Rebel sends the full conversation history to the AI model on every turn. When you correct inline, the model still sees:
- Your original (unclear) instruction
- Rebel’s wrong response — which may be long, with tool calls and reasoning
- Your “no, don’t do that” correction
All of that stays in the conversation permanently. The model has to work around the confusion, and its attention is split across contradictory signals. Over several turns, this adds up — your conversation gets noisier, and the quality of responses can drift.
What “Edit & Rerun” does differently
When you edit a previous message, Rebel removes everything after it — Rebel’s wrong response, your correction, all of it — and reruns with your improved instruction. The model sees only the cleaner prompt, with zero noise.
The shortcut: Press ⌘↑ (Mac) or Ctrl+↑ (Windows) when the composer is empty to jump to your last message and edit it.
When to use which approach
| Situation | Best approach |
|---|---|
| Rebel misunderstood your intent entirely |  Edit & Rerun — rewrite the instruction |
| Small tweak, response was mostly right |  Correct inline — “Actually, change X to Y” |
| Rebel took wrong actions (wrong tool, wrong file) | Edit & Rerun + check for side effects |
| Several rounds of back-and-forth confusion | Edit & Rerun from where it went off track |
One thing to watch out for
Edit & Rerun does not automatically undo real-world actions Rebel already took (sent emails, file changes, etc.). If Rebel already did something, you may need to ask it to reverse those separately.
TL;DR
Think of your conversation like a shared document. “Edit & Rerun” is like fixing a typo at the source. Correcting inline is like adding a footnote that says “ignore what I said above” — it works, but it makes the document messier for everyone reading it (including the AI).
Cleaner context → better results, especially in longer conversations. Give it a try next time something goes sideways! 