Backtracking after a `/review` causes a total loss of context
Resolved 💬 4 comments Opened Nov 21, 2025 by miraclebakelaser Closed Mar 6, 2026
What version of Codex is running?
0.61.0
What subscription do you have?
Pro
Which model were you using?
gpt-5.1-codex-max xhigh (but applies to all)
What platform is your computer?
Darwin 25.0.0 arm64 arm
What issue are you seeing?
After running /review in the middle of a conversation and resuming the conversation, backtracking (esc, esc) twice will cause the conversation to reset.
What steps can reproduce the bug?
- Start a conversation (say "hi")
- Run the
/reviewcommand (select any form of review) - After the review is completed, continue the conversation for 3 more turns.
Here's a sketch of a conversation:
user: hi
codex: hi how can i help you
/review
>> Code review started: changes against 'main' <<
user: ok thx
codex: sure
user: appreciate it
codex: np
user: bye
codex: 👍
- After saying bye, press
esc escto edit the previous message "user: bye". Delete the "bye" from the input box and then pressesc escto edit the "user: appreciate it" message.
You're then met with this message:
╭───────────────────────────────────────────────────────╮
│ >_ OpenAI Codex (v0.61.0) │
│ │
│ model: gpt-5.1-codex-max xhigh /model to change │
│ directory: ~/GitHub/codex │
╰───────────────────────────────────────────────────────╯
To get started, describe a task or try one of these commands:
/init - create an AGENTS.md file with instructions for Codex
/status - show current session configuration
/approvals - choose what Codex can do without approval
/model - choose what model and reasoning effort to use
/review - review any changes and find issues
›
100% context left · ? for shortcuts
The conversation is now reset and you cannot backtrack to any previous messages.
What is the expected behavior?
You should be able to backtrack to edit your previous message, and then backtrack again to edit another previous message, and so on.
Additional information
_No response_
This issue has 4 comments on GitHub. Read the full discussion on GitHub ↗