You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
{{ message }}
Repository navigation
Doesn't follow instructions and mixing up tasks. #7086
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.
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.
Thanks for the bug report. We'll share the feedback with the team responsible for training our codex models. We're tracking this feedback and similar reports so we can make future models more robust and efficient. And thanks for using the /feedback command. That additional information is useful.
Activity
github-actions commentedon Nov 21, 2025
Potential duplicates detected. Please review them and close your issue if it is a duplicate.
Checking unexpected file changesstep force resets all unrelated changes elsewhere #6187Powered by Codex Action
etraut-openai commentedon Nov 21, 2025
Thanks for the bug report. We'll share the feedback with the team responsible for training our codex models. We're tracking this feedback and similar reports so we can make future models more robust and efficient. And thanks for using the
/feedbackcommand. That additional information is useful.