What issue are you seeing?
I was developing a game on the cloud computer provided by OpenAI’s dot. After days of work, the environment appeared to change unexpectedly, and my project files were no longer accessible.
This happened in dot’s cloud computer, not in a separately selected Codex Cloud task. I did not request a new environment or initiate any deletion.
On October 2, 2026, between 10:23:12 and 10:23:47 UTC, access to the working environment failed. Subsequent checks showed a different machine identifier. The previous terminal and application windows were gone, and the project directory returned “No such file or directory.”
Repeated checks have not located the project in the accessible environment. I cannot currently determine whether the original storage still exists or whether recovery is possible. Permanent deletion has not been confirmed.
This is a serious reliability and data-persistence concern. Users can spend days developing software and configuring this cloud computer. If an environment change can suddenly make that work inaccessible, the persistence limits, backup requirements and recovery process need to be clearly disclosed before users rely on it.
Based on my experience, I consider this environment unsafe for important work when it holds the only copy. The conversation history does not replace the missing project files. Other files and configuration stored solely on the same machine may face the same risk; their loss has not been independently established in this incident.
Expected behavior:
- Project files remain accessible across normal use, or storage-lifecycle limits are prominently explained
- Users receive a clear warning and a practical export or backup option before any operation that could discard their working environment
- An unexpected environment change has an identifiable recovery path
Requested action:
- Investigate the environment change and locate the previous storage
- Preserve any available snapshots or saved state pending investigation
- Clarify persistence and recovery guarantees specifically for dot’s cloud computer
- Provide a reliable backup/export workflow and clear warnings about potential data loss
OpenAI Support has confirmed escalation to a human specialist. Recovery remains unconfirmed.
I’m reporting this here because of the relationship to the coding environment. If dot cloud-computer incidents belong elsewhere, please direct me to the responsible technical team or issue tracker. I can provide the conversation and machine identifiers privately to OpenAI Support.
What steps can reproduce the bug?
I do not have a reliable reproduction procedure because the environment change was unexpected. The observed sequence was:
- Use the cloud computer provided by dot within an ongoing conversation.
- Develop a game and save its files in the cloud computer’s workspace over several days.
- Continue working in the same conversation without requesting a new computer, resetting the environment or deleting the project.
- Access to the project unexpectedly fails. The previous terminal/application windows disappear.
- Check the machine identifier: it differs from the previous machine.
- Check the original project directory: it returns “No such file or directory.”
What is the expected behavior?
No response
Additional information
No response
What issue are you seeing?
I was developing a game on the cloud computer provided by OpenAI’s dot. After days of work, the environment appeared to change unexpectedly, and my project files were no longer accessible.
This happened in dot’s cloud computer, not in a separately selected Codex Cloud task. I did not request a new environment or initiate any deletion.
On October 2, 2026, between 10:23:12 and 10:23:47 UTC, access to the working environment failed. Subsequent checks showed a different machine identifier. The previous terminal and application windows were gone, and the project directory returned “No such file or directory.”
Repeated checks have not located the project in the accessible environment. I cannot currently determine whether the original storage still exists or whether recovery is possible. Permanent deletion has not been confirmed.
This is a serious reliability and data-persistence concern. Users can spend days developing software and configuring this cloud computer. If an environment change can suddenly make that work inaccessible, the persistence limits, backup requirements and recovery process need to be clearly disclosed before users rely on it.
Based on my experience, I consider this environment unsafe for important work when it holds the only copy. The conversation history does not replace the missing project files. Other files and configuration stored solely on the same machine may face the same risk; their loss has not been independently established in this incident.
Expected behavior:
Requested action:
OpenAI Support has confirmed escalation to a human specialist. Recovery remains unconfirmed.
I’m reporting this here because of the relationship to the coding environment. If dot cloud-computer incidents belong elsewhere, please direct me to the responsible technical team or issue tracker. I can provide the conversation and machine identifiers privately to OpenAI Support.
What steps can reproduce the bug?
I do not have a reliable reproduction procedure because the environment change was unexpected. The observed sequence was:
What is the expected behavior?
No response
Additional information
No response