Doesn't follow instructions and mixing up tasks.
What version of Codex is running?
codex-cli 0.61.0
What subscription do you have?
Pro
Which model were you using?
gpt-5.1-codex-max high
What platform is your computer?
WSL: Linux 6.6.87.2-microsoft-standard-WSL2 x86_64 x86_64 | On: Microsoft Windows NT 10.0.26200.0 x64
What issue are you seeing?
Told it we were working on a new issue: "Now to another issue."
I asked it to only trace the issue and not make changes: "Take a step back here! DO not make changes yet! Only trace and understand." It went ahead and made changes to an unrelated file, since it was mixing up previous tasks with this one.
What steps can reproduce the bug?
Uploaded thread: 019a9c99-ba8f-7010-a09d-e8b576f0893d
What is the expected behavior?
Follow instructions on not making changes, not mix up issues.
Additional information
After running /feedback I asked it to revert the changes it made when mixing up issues. It went ahead and deleted the whole file, then to try and recreate it. Here is it's reasoning: "I’m recreating the lost file with the latest working version of the OpenAIConnector class based on context clues and earlier code, preserving its
interface, methods for building requests, extracting text and sources, and switching to a named export as originally defined."
Recrating based on context clues usually ends up in a bad place.
This issue has 2 comments on GitHub. Read the full discussion on GitHub ↗